Method and system for compiling opendrive map

Abstract

The invention discloses a method for compiling an opendrive map, and the method comprises the following steps: receiving original data used for describing a road network structure, the original data comprising a discrete point set of a plurality of roads; generating and fitting a continuous reference line geometric model for each road based on the original data; performing alignment and smoothing processing on lane line data according to the reference line geometric model to generate a continuous lane geometric model and a width model; and integrating the reference line geometric model, the lane geometric model and the width model into a map file conforming to an opendrive standard. By fitting the continuous reference line and lane model from the discrete point set, sawteeth, jitter and inconsistency existing in original data are fundamentally eliminated, the generated map data better meet the dynamics requirements of vehicle motion, and a better and more stable path reference is provided for an automatic driving control module.

Description

Technical Field

[0001] This invention relates to the fields of map data processing and intelligent connected vehicle technology, specifically to a method and system for compiling OpenDrive maps. Background Technology

[0002] With the rapid development of autonomous driving and intelligent connected vehicles, high-precision maps are becoming increasingly important as a key component of vehicle environmental perception and decision-making. High-precision maps not only need to provide accurate road geometry information, but also need to contain rich semantic elements (such as lane lines, traffic signs, etc.) to serve the precise positioning and route planning of vehicles.

[0003] OpenDrive, as an open and standardized static scene data format, has become the de facto standard for describing high-precision road networks. Through reference lines, lane models, and rich attribute data, it can accurately describe complex road geometry and topology. Therefore, the efficient and automatic generation of map data conforming to the OpenDrive standard is crucial for promoting the large-scale application of autonomous driving technology.

[0004] Currently, common data sources for generating OpenDrive maps include point cloud data collected by vehicle sensors, image data, and discrete road point sets exported by professional surveying software. However, existing technologies have significant shortcomings in processing this raw data:

[0005] Generating OpenDrive files directly from a discrete set of points often results in reference lines being composed of numerous tiny straight or curved segments, leading to discontinuous curvature and an inability to smoothly reflect the true geometry of the road. This can negatively impact the stability and comfort of autonomous vehicle control algorithms.

[0006] Lane line data is usually closely related to road reference lines. If the reference lines themselves are not smooth, directly generating lane lines by offsetting them will cause distortion and jaggedness in the lane geometry model, destroying the parallelism and smooth transition between lanes, making it difficult to meet the stringent requirements of high-precision maps.

[0007] Many methods rely on extensive manual intervention and post-editing to correct geometric models, or the generated files deviate from the best practices of the OpenDrive standard in terms of geometric description and parameterization, resulting in poor data compatibility and low generation efficiency. Summary of the Invention

[0008] To overcome the above-mentioned technical defects, the present invention provides a method and system for compiling OpenDrive maps.

[0009] To solve the above problems, the present invention is implemented according to the following technical solution:

[0010] In a first aspect, the present invention provides a method for compiling an OpenDrive map, comprising the following steps: receiving raw data for describing the road network structure, the raw data including discrete point sets of multiple roads; generating and fitting a continuous reference line geometric model for each road based on the raw data; aligning and smoothing lane line data according to the reference line geometric model to generate a continuous lane geometric model and width model; and integrating the reference line geometric model, lane geometric model and width model into a map file conforming to the OpenDrive standard.

[0011] In conjunction with the first aspect, the present invention provides a first specific implementation of the first aspect. Specifically, the step of generating and fitting a continuous reference line geometric model for each road based on the original data includes the following steps: generating a discrete point set for each road reference line and establishing the connection relationship between adjacent reference lines; calculating the geometric tortuosity of each reference line, wherein the geometric tortuosity is defined as the ratio of the total path length of the reference line point set to the Euclidean distance between the first and last points; and fitting all reference lines using a cubic spline algorithm with endpoint tangent direction constraints according to the geometric tortuosity from low to high to generate a continuous reference line geometric model.

[0012] In conjunction with the first aspect, the present invention provides a second specific implementation of the first aspect. Specifically, when the cubic spline algorithm with endpoint tangent direction constraints is used for fitting, the logic for determining the endpoint tangent direction is as follows: the tangent direction data of the starting point is preferentially inherited from the endpoint tangent direction of its predecessor reference line; if the predecessor reference line is not fitted, it is calculated based on the vector of the starting segment of the current reference line; the tangent direction data of the endpoint is preferentially inherited from the starting point tangent direction of its successor reference line; if the successor reference line is not fitted, it is calculated based on the vector of the ending segment of the current reference line.

[0013] In conjunction with the first aspect, the present invention provides a third specific implementation of the first aspect. Specifically, the alignment and smoothing of lane line data according to the reference line geometric model includes: comparing the geometric length of each lane line with the length of the corresponding reference line geometric model; and adjusting the geometry of all lane lines by inserting or deleting lane line points to make their length consistent with the length of the reference line geometric model.

[0014] In conjunction with the first aspect, the present invention provides a fourth specific implementation of the first aspect. Specifically, the width model construction includes: calculating the distance between the left and right boundaries of the lane along the reference line position sequence to obtain a discrete width sampling sequence; performing digital filtering on the discrete width sampling sequence to smooth data fluctuations; and fitting the smoothed width sequence as a piecewise cubic polynomial function with respect to the reference line position.

[0015] Secondly, the present invention also provides a system for compiling OpenDrive maps, comprising: an input module for receiving raw data describing the road network structure, the raw data including discrete point sets of multiple roads; a reference line processing module for generating and fitting a continuous reference line geometric model for each road based on the raw data; a lane processing module for aligning and smoothing lane line data according to the reference line geometric model to generate a continuous lane geometric model and a width model; and an output module for integrating the reference line geometric model, the lane geometric model, and the width model into a map file conforming to the OpenDrive standard.

[0016] In conjunction with the second aspect, the present invention provides a first specific implementation of the second aspect. Specifically, the operations performed by the reference line processing module include: generating a discrete set of reference lines for each road and establishing connection relationships between adjacent reference lines; calculating the geometric tortuosity of each reference line, wherein the geometric tortuosity is defined as the ratio of the total path length of the reference line point set to the Euclidean distance between the first and last points; and fitting all reference lines using a cubic spline algorithm with endpoint tangent direction constraints according to the geometric tortuosity from low to high, so as to generate a continuous reference line geometric model.

[0017] In conjunction with the second aspect, the present invention provides a second specific implementation of the second aspect. Specifically, when the cubic spline algorithm with endpoint tangent direction constraints is used for fitting, the logic for determining the endpoint tangent direction is as follows: the tangent direction data of the starting point is preferentially inherited from the endpoint tangent direction of its predecessor reference line; if the predecessor reference line is not fitted, it is calculated based on the vector of the starting segment of the current reference line; the tangent direction data of the endpoint is preferentially inherited from the starting point tangent direction of its successor reference line; if the successor reference line is not fitted, it is calculated based on the vector of the ending segment of the current reference line.

[0018] In conjunction with the second aspect, the present invention provides a third specific implementation of the second aspect. Specifically, the lane line alignment operation performed by the lane processing module includes: comparing the geometric length of each lane line with the length of the corresponding reference line geometric model; and adjusting the geometry of all lane lines by inserting or deleting lane line points so that their lengths are consistent with the lengths of the reference line geometric model.

[0019] In conjunction with the second aspect, the present invention provides a fourth specific implementation of the second aspect. Specifically, the lane width model construction operation performed by the lane processing module includes: calculating the distance between the left and right boundaries of the lane along the reference line position sequence to obtain a discrete width sampling sequence; performing digital filtering on the discrete width sampling sequence to smooth data fluctuations; and fitting the smoothed width sequence as a piecewise cubic polynomial function with respect to the reference line position.

[0020] Compared with the prior art, the beneficial effects of the present invention are:

[0021] By fitting continuous reference lines and lane models from discrete point sets, the jaggedness, jitter, and discontinuity in the original data are fundamentally eliminated. The generated map data better conforms to the dynamic requirements of vehicle motion, providing a higher quality and more stable path reference for the autonomous driving control module. The unified conversion and output of raw data from diverse sources and in different formats into the internationally recognized OpenDRIVE standard format greatly enhances the interoperability and exchangeability of map data, facilitating seamless use across different simulation platforms and autonomous driving systems. Modeling lanes based on reference lines ensures the geometric correlation and topological correctness between lanes and roads, and between adjacent lanes, effectively avoiding problems such as lane overlap, breaks, or logical connection errors that may occur in traditional methods. Attached Figure Description

[0022] The specific embodiments of the present invention will be further described in detail below with reference to the accompanying drawings, wherein:

[0023] Figure 1 This is a flowchart of a method for compiling an OpenDrive map according to the present invention. Detailed Implementation

[0024] The preferred embodiments of the present invention will be described below with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are for illustration and explanation only and are not intended to limit the present invention.

[0025] like Figure 1 As shown, this invention provides a method and system for compiling OpenDrive maps.

[0026] Example 1

[0027] A method for compiling an OpenDrive map includes the following steps: receiving raw data for describing the road network structure, the raw data including discrete point sets of multiple roads; generating and fitting a continuous reference line geometric model for each road based on the raw data; aligning and smoothing lane line data according to the reference line geometric model to generate a continuous lane geometric model and width model; and integrating the reference line geometric model, lane geometric model, and width model into a map file conforming to the OpenDrive standard.

[0028] In a preferred embodiment, raw data describing the road network structure is received, the raw data comprising discrete point sets of multiple roads. Each road's discrete point set at least defines the approximate orientation of a reference line (typically the road centerline or a curb) for that road. The raw data may further include information such as lane line positions and lane widths associated with these discrete points.

[0029] In a preferred embodiment, a continuous reference line geometry model is generated and fitted for each road based on the original data.

[0030] Specifically, the discrete point set of a road is connected sequentially to form an initial, potentially non-smooth, polygonal line as the initial centerline. A curve fitting algorithm (such as cubic spline curve, polynomial fitting, or Bézier curve) is used to fit the initial centerline, generating a smooth and continuous curve. The purpose is to eliminate noise and jagged edges in the original data, ensuring that the reference line is geometrically continuous at least in terms of tangent, and preferably in terms of curvature. The fitted smooth curve is parameterized as a function of arc length s, i.e., (x(s), y(s), h(s)), where x(s) and y(s) are coordinates, and h(s) is the heading angle. This parameterized curve is the geometric model of the road's reference line.

[0031] In a preferred embodiment, lane line data is aligned and smoothed according to the reference line geometry model to generate a continuous lane geometry model and width model.

[0032] Specifically, discrete points related to lanes in the original data (such as lane boundary points and lane center points) are mapped onto the reference line generated in step S2. For each lane discrete point, its nearest projection point on the reference line is calculated, and the arc length coordinates s of that projection point and the lateral offset t between the lane point and the reference line are recorded. For each lane, based on the (s,t) pairs obtained after mapping all its discrete points, a discrete sequence describing the width or offset of the lane at different positions s on the reference line is formed. The obtained discrete width sequence is smoothed and fitted (e.g., using a polynomial or spline function) to generate a continuous lane width function width(s) with the arc length s of the reference line as the independent variable. This function defines the width model of the lane relative to the reference line. Based on the reference line geometric model (x(s), y(s), h(s)) and the lane width model width(s), the continuous geometric model of the lane centerline or boundary can be obtained through lateral offset calculation. The specific formula is:

[0033] X lanes (s)=x(s)+t(s)*sin(h(s));

[0034] Y lanes (s)=y(s)+t(s)*cos(h(s));

[0035] Where t(s) is the lateral offset derived from the width model, x(s) and y(s) are the coordinates, and h(s) is the heading angle. The lane lines generated in this way are also smooth and continuous.

[0036] In a preferred embodiment, the reference line geometry model, lane geometry model, and width model are integrated into a map file conforming to the OpenDrive standard.

[0037] Specifically, the generated reference line geometry model for each road uses geometric elements supported by OpenDrive. <line> , <spiral> , <arc> , <poly3>Describe the information in segments (e.g., etc.) and fill in the blanks. <road>Below <planview>Part. In the corresponding <road>Create <lanes>Segment. Based on the generated lane width model, in <lanesection>The middle is for each lane <lane>definition <width>Records. These <width>The record is a combination of arc length s and polynomial coefficients a, b, c, d, which precisely describes width(s) = a + b*s + c*s 2 +d*s 3 Based on the connectivity between roads in the original road network, opendrive uses ( <link> ( <predecessor> , <successor>)and <junction>Elements are used to establish topological connections between roads. All structured data is assembled into a complete .xodr XML file that conforms to the OpenDrive Schema specification.

[0038] In a preferred embodiment, generating and fitting a continuous reference line geometric model for each road based on the original data specifically includes the following steps: generating a discrete point set for each road reference line and establishing the connection relationship between adjacent reference lines; calculating the geometric tortuosity of each reference line, wherein the geometric tortuosity is defined as the ratio of the total path length of the reference line point set to the Euclidean distance between the first and last points; and fitting all reference lines using a cubic spline algorithm with endpoint tangent direction constraints according to the geometric tortuosity from low to high, so as to generate a continuous reference line geometric model.

[0039] Specifically, the generated and fitted reference line geometric model was further refined and enhanced. Its core lies in evaluating the road geometric complexity and determining the optimal fitting order, while introducing tangent constraints at connections to ensure a smooth transition across the entire road network, especially at road connection points. This preferred solution includes the following steps:

[0040] Extract or generate a discrete point set P for the reference line of each road from the original data. i Analyze the road network topology to identify all adjacent road reference lines. For each pair of adjacent reference lines, record their spatial relationship at the connection point. For example, the end point P_end_A of road A is connected to the start point P_start_B of road B.

[0041] Calculate a quantitative geometric tortuosity γ for the discrete point set of the reference line for each road. Geometric tortuosity γ is defined as the total length L of the polygonal segment formed by all consecutive discrete points of the reference line. total The Euclidean distance L between its first and last points euclidean The ratio of λ to L. Its calculation formula is: λ = L total / L euclidean When λ≈1, it means that the reference line is almost a straight line with a simple geometry and low tortuosity. When λ>>1, it means that the reference line has a high degree of curvature (such as a mountain road or an S-shaped bend), a complex geometry, and high tortuosity.

[0042] Based on the calculated geometric tortuosity, all reference lines are sorted in ascending order of γ value (i.e., from the "straightest" road to the "most curved" road). This order ensures that roads with simple geometry are processed first during the fitting process.

[0043] In a preferred embodiment, when the cubic spline algorithm with endpoint tangent direction constraints is used for fitting, the logic for determining the endpoint tangent direction is as follows: the tangent direction data of the starting point is preferentially inherited from the endpoint tangent direction of its predecessor reference line; if the predecessor reference line is not fitted, it is calculated based on the vector of the starting segment of the current reference line; the tangent direction data of the endpoint is preferentially inherited from the starting point tangent direction of its successor reference line; if the successor reference line is not fitted, it is calculated based on the vector of the ending segment of the current reference line.

[0044] Specifically, cubic spline fitting with endpoint tangent direction constraints: Following the above order, curve fitting is performed on each reference line sequentially. During the fitting process, a cubic spline algorithm with endpoint tangent direction constraints is used. The specific constraint rules are as follows:

[0045] Determining the starting tangent direction: If the current road has a fitted predecessor road, the starting tangent direction is directly inherited from the ending tangent direction of the predecessor road. If the predecessor road has not yet been fitted (e.g., it is the first road, or it has high tortuosity and is ranked later), it is calculated based on the local vector of the starting segment of the current reference line (e.g., the vector direction from the first two discrete points P0 to P1).

[0046] The tangent direction at the endpoint is determined as follows: If a fitted successor road exists for the current road, the tangent direction at its endpoint is directly inherited from the tangent direction at the starting point of that successor road. If no successor road has been fitted, the tangent direction is determined based on the local vector of the end segment of the current reference line (e.g., taking the last two discrete points P). n-1 To P n The vector direction is used for calculation.

[0047] For isolated roads or the first fitted road: the tangent directions of its starting and ending points can be calculated through local point sets (e.g., taking the directions of the first few points as the starting tangent and the directions of the last few points as the ending tangent).

[0048] For roads connected to already fitted roads: Suppose the starting point of the road B to be fitted is connected to the ending point of the already fitted road A. Then, when fitting road B, the tangent direction at its starting point will be forcibly constrained to be consistent with the tangent direction at the ending point of road A.

[0049] Generate a continuous reference line geometric model: The curve fitted by this algorithm is not only smooth and continuous, but more importantly, it ensures tangent continuity at road junctions (intersections), eliminating sharp angles or abrupt changes in direction that may exist at junctions, thereby greatly improving the geometric quality and driving comfort of the generated road network.

[0050] In a preferred embodiment, the alignment and smoothing of lane line data according to the reference line geometry model specifically includes: comparing the geometric length of each lane line with the length of the corresponding reference line geometry model; and adjusting the geometry of all lane lines by inserting or deleting lane line points to make their length consistent with the length of the reference line geometry model.

[0051] Specifically, to ensure a strict correspondence between the lane geometry model and the reference line geometry model in the parameter space, and to avoid length mismatch issues caused by errors in the original data acquisition, this solution introduces a standardization step for lane line geometric lengths before data mapping. This includes:

[0052] Before mapping the lane line data, the original geometric length L of each lane line to be processed (whether it is the lane centerline or the boundary line) is first calculated. lane And the length L of the generated, corresponding reference line geometry model. ref A comparison is then made. Based on the comparison results, the discrete point set of the lane lines is geometrically adjusted to ensure that its length is consistent with the reference line model.

[0053] If L lane Significantly shorter than L ref This indicates that the original lane line data may be missing or incomplete. The system extends it by inserting new points into the lane line shape point sequence. The insertion strategy can be to increase the number of points evenly and proportionally, or to insert more points in areas of high curvature based on the curvature of the reference line to better fit the shape of the reference line.

[0054] If L lane Significantly longer than L ref This indicates that the original lane line data may contain redundancy or noise. The system shortens it by deleting redundant points, typically in sections with low curvature and relatively straight lines, to eliminate excess length while preserving key geometric features. Through the above insertion or deletion operations, the geometric length of all lane lines is adjusted to match the length L of the corresponding reference line geometric model. ref Consistent.

[0055] The operation is performed on a discrete set of lane line points after length normalization. Each lane shape point is mapped onto a reference line, and the arc length coordinates s and lateral offset t of its projection point are calculated. For each lane, based on the (s, t) pairs obtained from all its normalized shape point mappings, a discrete sequence describing the width or offset of the lane at different positions s on the reference line is formed. The resulting discrete width sequence is smoothly fitted to generate a continuous lane width function width(s) with the reference line arc length s as the independent variable, i.e., the lane width model. Based on the reference line geometric model and the lane width model, a continuous and smooth geometric model of the lane centerline or boundary is obtained through lateral offset calculation.

[0056] This preferred embodiment fundamentally solves the problem of mismatch between reference line and lane line length caused by inconsistent data sources, establishing an accurate and consistent length benchmark for subsequent projection mapping. Length consistency ensures that each point on the lane line can find a more reasonable and accurate projection point s on the reference line. If the lane line is too short, its end point cannot be projected onto the end of the reference line, resulting in data loss; if it is too long, the extra points at the end may be incorrectly projected, introducing noise. Standardization effectively avoids both of these situations. Length consistency generates a range from s=0 to s=L. ref A complete, continuous, and uninterrupted lane width modulus (width(s)) is a prerequisite. This is crucial for the correct generation of lane width records in OpenDrive.

[0057] In a preferred embodiment, the width model construction specifically includes: calculating the distance between the left and right boundaries of the lane along the reference line position sequence to obtain a discrete width sampling sequence; performing digital filtering on the discrete width sampling sequence to smooth data fluctuations; and fitting the smoothed width sequence as a piecewise cubic polynomial function with respect to the reference line position.

[0058] Specifically, to ensure a strict correspondence between the lane geometry model and the reference line geometry model in the parameter space, and to avoid length mismatch issues caused by errors in the original data acquisition, this solution introduces a standardization step for lane line geometric lengths before data mapping. This includes:

[0059] For each lane, based on the (s, t) pairs obtained from all its standardized shape point mappings, a discrete sequence describing the width or offset of that lane at different positions s on the reference line is formed. This discrete, potentially noisy, width sequence is then used to construct a continuous, smooth, and accurate width model, width(si). Specifically, this includes the following sub-steps:

[0060] The position sequence along the reference line (usually a series of arc length coordinates s generated with a fixed or adaptive step size from 0 to the total length L of the reference line) i ), calculate the lateral offset t of the left and right boundaries of the lane relative to the reference line. left (s i ) and t right (s i Then, by calculating width(s) i )=t right (s i )-t left (s i This yields a series of discrete width sample values ​​{s}. i w i } constitutes a discrete-width sampling sequence. The resulting discrete-width sampling sequence {s} i w i Digital filtering (e.g., using a moving average filter, Gaussian filter, or Savitzky-Golay filter) is performed to smooth out high-frequency components in the data caused by measurement noise or minor fluctuations, while preserving the overall trend of width variation. This step effectively eliminates data fluctuations, providing a clean and smooth data foundation for subsequent function fitting. The digitally filtered and smoothed width sequence is then fitted as a piecewise cubic polynomial function about the reference line position s. Based on the geometric characteristics of the reference line (e.g., points of curvature change) or the drastic change in width itself, the entire arc length range [0, L] is divided into several continuous intervals. In each interval [s... k s k+1 Within [the context], the width function width(s) is derived from a cubic polynomial w k (s)=a k +b k *(ss k )+c k *(ss k ) 2 +d k *(ss k ) 3 To describe this, at the segmentation points, adjacent polynomials are required to maintain consistency in function values ​​(zero-order continuity) and first derivatives (first-order continuity, i.e., smooth slope change) to ensure that the generated width model is smooth overall. Based on the reference line geometry model and the constructed smooth piecewise cubic polynomial width model, a continuous and smooth geometry model of the lane centerline or boundary is obtained through lateral offset calculation.

[0061] This preferred embodiment uses piecewise cubic polynomials to accurately fit various complex width changes, such as gradual widening, narrowing, or S-shaped lane changes, demonstrating a much stronger expressive power than simple linear models. Through digital filtering preprocessing and continuity constraints at segmentation points, the final width model is not only smooth but also exhibits a continuous first derivative (i.e., the rate of width change), consistent with the physical characteristics of gradual lane width changes in the real world. In the OpenDrive standard, lane width is defined using piecewise cubic polynomials. This method directly generates data structures conforming to this standard, eliminating the need for additional conversion steps during compilation and ensuring the standardization and compatibility of the output format.

[0062] Example 2

[0063] A system for compiling OpenDrive maps includes: an input module for receiving raw data describing a road network structure, the raw data including discrete point sets of multiple roads; a reference line processing module for generating and fitting a continuous reference line geometric model for each road based on the raw data; a lane processing module for aligning and smoothing lane line data according to the reference line geometric model to generate a continuous lane geometric model and a width model; and an output module for integrating the reference line geometric model, the lane geometric model, and the width model into a map file conforming to the OpenDrive standard.

[0064] In a preferred embodiment, the operations performed by the reference line processing module include: generating a discrete set of reference line points for each road and establishing connection relationships between adjacent reference lines; calculating the geometric tortuosity of each reference line, wherein the geometric tortuosity is defined as the ratio of the total path length of the reference line point set to the Euclidean distance between the first and last points; and fitting all reference lines to generate a continuous reference line geometric model by using a cubic spline algorithm with endpoint tangent direction constraints, according to the geometric tortuosity from low to high.

[0065] In a preferred embodiment, when the cubic spline algorithm with endpoint tangent direction constraints is used for fitting, the logic for determining the endpoint tangent direction is as follows: the tangent direction data of the starting point is preferentially inherited from the endpoint tangent direction of its predecessor reference line; if the predecessor reference line is not fitted, it is calculated based on the vector of the starting segment of the current reference line; the tangent direction data of the endpoint is preferentially inherited from the starting point tangent direction of its successor reference line; if the successor reference line is not fitted, it is calculated based on the vector of the ending segment of the current reference line.

[0066] In a preferred embodiment, the lane alignment operation performed by the lane processing module includes: comparing the geometric length of each lane line with the length of the corresponding reference line geometric model; and adjusting the geometry of all lane lines by inserting or deleting lane line points to make their length consistent with the length of the reference line geometric model.

[0067] In a preferred embodiment, the lane width model construction operation performed by the lane processing module includes: calculating the distance between the left and right boundaries of the lane along the reference line position sequence to obtain a discrete width sampling sequence; performing digital filtering on the discrete width sampling sequence to smooth data fluctuations; and fitting the smoothed width sequence as a piecewise cubic polynomial function with respect to the reference line position.

[0068] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Therefore, any modifications, equivalent changes, and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.< / junction> < / successor> < / predecessor> < / width> < / width> < / lane> < / lanesection> < / lanes> < / road> < / planview> < / road> < / arc> < / spiral> < / line>

Claims

1. A method of compiling an opendrive map, characterized by, Includes the following steps: Receive raw data for describing the road network structure, the raw data including discrete point sets of multiple roads; Based on the original data, a continuous reference line geometric model is generated and fitted for each road; The lane line data is aligned and smoothed based on the reference line geometry model to generate a continuous lane geometry model and width model. The reference line geometry model, lane geometry model, and width model are integrated into a map file that conforms to the OpenDrive standard.

2. The method of compiling an opendrive map according to claim 1, wherein, The process of generating and fitting a continuous reference line geometric model for each road based on the original data specifically includes the following steps: Generate a set of discrete reference points for each road and establish the connection relationship between adjacent reference lines; Calculate the geometric tortuosity of each reference line, whereby the geometric tortuosity is defined as the ratio of the total path length of the reference line point set to the Euclidean distance between the first and last points; Based on the geometric tortuosity from low to high, a cubic spline algorithm with endpoint tangent direction constraints is used to fit all reference lines to generate a continuous reference line geometric model.

3. The method of compiling an opendrive map according to claim 2, wherein, When using the cubic spline algorithm with endpoint tangent direction constraints for fitting, the logic for determining the endpoint tangent direction is as follows: The tangent direction data of the starting point is preferentially inherited from the end point tangent direction of its predecessor reference line. If the predecessor reference line is not fitted, it is calculated based on the vector of the starting segment of the current reference line. The tangent direction data at the endpoint is preferentially inherited from the starting tangent direction of its successor reference line. If the successor reference line is not fitted, it is calculated based on the vector of the ending segment of the current reference line.

4. The method of compiling an opendrive map according to claim 1, wherein, The alignment and smoothing of lane line data based on the reference line geometric model specifically includes: Compare the geometric length of each lane line with the length of the corresponding reference line geometric model; By inserting or deleting lane line points, all lane lines are geometrically adjusted to match the length of the reference line geometry model.

5. The method according to claim 1, characterized in that, The construction of the width model specifically includes: Along the reference line position sequence, calculate the distance between the left and right boundaries of the lane to obtain a discrete width sampling sequence; The discrete width sampling sequence is digitally filtered to smooth data fluctuations; The smoothed width sequence is fitted as a piecewise cubic polynomial function with respect to the position of the reference line.

6. A system for compiling OpenDrive maps, characterized in that, include: The input module is used to receive raw data for describing the road network structure, the raw data including a discrete point set of multiple roads; The reference line processing module is used to generate and fit a continuous reference line geometric model for each road based on the original data. The lane processing module is used to align and smooth the lane line data according to the reference line geometry model to generate a continuous lane geometry model and width model. The output module is used to integrate the reference line geometry model, lane geometry model, and width model into a map file that conforms to the OpenDrive standard.

7. A system for compiling OpenDrive maps according to claim 6, characterized in that, The operations performed by the reference line processing module include: Generate a set of discrete reference points for each road and establish the connection relationship between adjacent reference lines; Calculate the geometric tortuosity of each reference line, whereby the geometric tortuosity is defined as the ratio of the total path length of the reference line point set to the Euclidean distance between the first and last points; Based on the geometric tortuosity from low to high, a cubic spline algorithm with endpoint tangent direction constraints is used to fit all reference lines to generate a continuous reference line geometric model.

8. A system for compiling OpenDrive maps according to claim 7, characterized in that, When using the cubic spline algorithm with endpoint tangent direction constraints for fitting, the logic for determining the endpoint tangent direction is as follows: The tangent direction data of the starting point is preferentially inherited from the end point tangent direction of its predecessor reference line. If the predecessor reference line is not fitted, it is calculated based on the vector of the starting segment of the current reference line. The tangent direction data at the endpoint is preferentially inherited from the starting tangent direction of its successor reference line. If the successor reference line is not fitted, it is calculated based on the vector of the ending segment of the current reference line.

9. A system for compiling OpenDrive maps according to claim 6, characterized in that, The lane alignment operation performed by the lane processing module includes: Compare the geometric length of each lane line with the length of the corresponding reference line geometric model; By inserting or deleting lane line points, all lane lines are geometrically adjusted to match the length of the reference line geometry model.

10. A system for compiling OpenDrive maps according to claim 6, characterized in that, The lane width model construction operation performed by the lane processing module includes: Along the reference line position sequence, calculate the distance between the left and right boundaries of the lane to obtain a discrete width sampling sequence; The discrete width sampling sequence is digitally filtered to smooth data fluctuations; The smoothed width sequence is fitted as a piecewise cubic polynomial function with respect to the position of the reference line.

Images