Road trajectory generation method, device and vehicle

Abstract

The application provides a road trajectory generation method and device and a vehicle, and relates to the technical field of intelligent driving control. Road shape points and visual lane line point sequences in a navigation map can be acquired; the road shape points and the visual lane line point sequences are unified to a vehicle coordinate system; the positions of the road shape points in the vehicle coordinate system are corrected based on the visual lane line point sequences; and a road trajectory in front of the vehicle is generated based on the corrected road shape points. In this way, in a scene where visual lane line detection is limited or the effective length is insufficient, a continuous and accurate road trajectory in front of the vehicle can be generated by using road shape points in a standard navigation-level map, a reference is provided for vehicle lateral control, trajectory deviation and swing phenomena are reduced, and dependence on a high-precision map is reduced.

Description

Technical Field

[0001] This application relates to the field of intelligent driving control technology, and in particular to a road trajectory generation method, device and vehicle. Background Technology

[0002] In driver assistance systems, lateral control typically requires obtaining the road trajectory ahead of the vehicle as a reference to ensure stable driving along the desired path. This is especially crucial in complex road structures such as intersections, ramps, and sharp curves, where the continuity and accuracy of the road trajectory ahead are of paramount importance.

[0003] In related technologies, visually perceived lane line information is typically used for lateral control to obtain the trajectory of the road ahead. On regular roads, visual lane lines can provide effective lane-level references. However, in scenarios such as intersections, ramps, and sharp curves, visual lane lines often suffer from missing lines, complex structures, or insufficient effective visible length, making it difficult to obtain continuous and stable road references relying solely on visual lane lines. To compensate for the failure of visual lane lines, some related technologies use high-precision maps; however, the acquisition and maintenance costs of high-precision maps are high, and they are often difficult to deploy directly on low-computing-power platforms.

[0004] Therefore, given the limited reliability of visual lane line detection and the difficulty for low-computing-power platforms to utilize high-precision maps, how to generate continuous and reliable forward road trajectories as a reference for lateral control, thereby reducing vehicle trajectory deviation and "swaying" phenomena, has become a pressing technical problem to be solved in related technologies. Summary of the Invention

[0005] This application provides a method, apparatus, and vehicle for generating road trajectories.

[0006] Firstly, a method for generating road trajectories is provided, comprising: acquiring road shape points and visual lane line point sequences from a navigation map, wherein the visual lane line point sequence is a sequence of points on the lane lines perceived by the vehicle; unifying the road shape points and visual lane line point sequences into the vehicle coordinate system; correcting the position of the road shape points in the vehicle coordinate system based on the visual lane line point sequence; and generating the road trajectory ahead of the vehicle based on the corrected road shape points.

[0007] In one embodiment, the position of road shape points in the vehicle coordinate system is corrected based on the visual lane line point sequence, including: establishing a correspondence between road shape points and the visual lane line point sequence in the vehicle coordinate system; calculating spatial transformation parameters based on the correspondence; and transforming the position of road shape points in the vehicle coordinate system based on the spatial transformation parameters to obtain the corrected road shape points.

[0008] In one embodiment, obtaining a visual lane line point sequence includes: selecting a target lane line from lane lines perceived by the vehicle's visual perception according to the vehicle's direction of travel; and performing discrete sampling within a preset range of the target lane line to obtain a visual lane line point sequence.

[0009] In one embodiment, selecting a target lane line from the lane lines perceived by the vehicle's vision according to the vehicle's direction of travel includes: selecting the left lane line from the lane lines perceived by the vehicle's vision as the target lane line when the vehicle's direction of travel is a right turn; selecting the right lane line from the lane lines perceived by the vehicle's vision as the target lane line when the vehicle's direction of travel is a left turn; and selecting the center line of the lane lines perceived by the vehicle's vision as the target lane line when the vehicle's direction of travel is straight.

[0010] In one embodiment, unifying the road shape points and visual lane line point sequences into the vehicle coordinate system includes: compensating the road shape points from historical moments to the vehicle coordinate system at the current moment based on the vehicle's own speed and heading angular velocity; and compensating the visual lane line point sequences from historical moments to the vehicle coordinate system at the current moment based on the vehicle's own speed and heading angular velocity.

[0011] In one embodiment, the method further includes: integrating the vehicle's own speed provided by the inertial measurement unit over time to estimate the longitudinal and lateral displacements of the vehicle from the moment the data from the latest received data from the global navigation satellite system to the current moment, and calculating the vehicle's current position as the origin of the vehicle coordinate system; using the attitude angle output by the latest data from the global navigation satellite system as the initial reference, recursively obtaining the vehicle's heading angle at the current moment by fusing the heading angular velocity from the inertial measurement unit, so as to use the vehicle's heading as the x-axis direction of the vehicle coordinate system; and constructing the vehicle coordinate system at the current moment based on the origin of the vehicle coordinate system and the x-axis direction of the vehicle coordinate system.

[0012] In one embodiment, the heading angle of the vehicle at the current moment is obtained by recursively calculating the heading angle of the vehicle using the latest attitude angle output by the Global Navigation Satellite System as the initial reference and fusing the heading angular velocity of the inertial measurement unit, so as to use the heading of the vehicle as the x-axis direction of the vehicle coordinate system. This includes: using Kalman filtering with the latest attitude angle output by the Global Navigation Satellite System as the observation and the heading angular velocity of the inertial measurement unit as the state recursion input to perform optimal estimation of the attitude angle state, so as to obtain the heading angle of the vehicle at the current moment and use the heading of the vehicle as the x-axis direction of the vehicle coordinate system.

[0013] In one embodiment, establishing the correspondence between road shape points and the visual lane line point sequence in the vehicle coordinate system includes: calculating the comprehensive distance metric between each road shape point and each point in the visual lane line point sequence in the vehicle coordinate system, wherein the comprehensive distance metric is obtained by weighted fusion of Euclidean distance, curvature similarity and radian angle difference; using the comprehensive distance metric as the matching cost, the dynamic time warping algorithm is used to solve the correspondence with the minimum global matching cost through dynamic programming.

[0014] In one embodiment, calculating spatial transformation parameters based on correspondence includes: calculating the optimal rotation matrix from the set of road shape points to the corresponding set of visual lane line point sequences using singular value decomposition based on the correspondence, as the spatial transformation parameters.

[0015] In one embodiment, generating a road trajectory ahead of a vehicle based on corrected road shape points includes: removing outliers from the corrected road shape points according to curvature continuity and positional trends; performing linear interpolation on sparse shape points whose spacing is greater than a set spacing threshold and whose curvature of the line connecting adjacent shape points is less than a set curvature threshold; and performing curve fitting based on the road shape points after outlier removal and linear interpolation to obtain the road trajectory.

[0016] In one embodiment, curve fitting is performed on the road shape points after outlier removal and linear interpolation to obtain the road trajectory, including: fitting the road shape points after outlier removal and linear interpolation with cubic Bézier curves or cubic spline curves to obtain the road trajectory.

[0017] Secondly, a road trajectory generation device is provided, comprising: an acquisition module for acquiring road shape points and visual lane line point sequences from a navigation map, wherein the visual lane line point sequence is a sequence of points on the lane lines perceived by the vehicle; a processing module for unifying the road shape points and visual lane line point sequences into a vehicle coordinate system; correcting the position of the road shape points in the vehicle coordinate system based on the visual lane line point sequence; and generating a road trajectory ahead of the vehicle based on the corrected road shape points.

[0018] In one embodiment, the processing module is specifically used to establish a correspondence between road shape points and visual lane line point sequences in the vehicle coordinate system; calculate spatial transformation parameters based on the correspondence; and transform the position of the road shape points in the vehicle coordinate system based on the spatial transformation parameters to obtain corrected road shape points.

[0019] In one embodiment, the acquisition module is specifically used to select a target lane line from the lane lines perceived by the vehicle's vision according to the vehicle's direction of travel; and to perform discrete sampling within a preset range of the target lane line to obtain a visual lane line point sequence.

[0020] In one embodiment, the acquisition module is specifically configured to select the left lane line of the lane lines perceived by the vehicle as the target lane line when the vehicle's direction of travel is right turn; select the right lane line of the lane lines perceived by the vehicle as the target lane line when the vehicle's direction of travel is left turn; and select the center line of the lane lines perceived by the vehicle as the target lane line when the vehicle's direction of travel is straight.

[0021] In one embodiment, the processing module is specifically used to compensate the road shape points from the historical time to the vehicle coordinate system at the current time based on the vehicle's own speed and heading angular velocity, and to compensate the visual lane line point sequence from the historical time to the vehicle coordinate system at the current time based on the vehicle's own speed and heading angular velocity.

[0022] In one embodiment, the processing module is further configured to perform time integration on the vehicle speed provided by the inertial measurement unit, estimate the longitudinal and lateral displacements of the vehicle from the moment the latest data from the global navigation satellite system was received to the current moment, and calculate the vehicle position at the current moment as the origin of the vehicle coordinate system; using the attitude angle output by the latest GPS system as the initial reference, and recursively obtaining the heading angle of the vehicle at the current moment by fusing the heading angular velocity of the inertial measurement unit, so as to use the heading of the vehicle as the x-axis direction of the vehicle coordinate system; and construct the vehicle coordinate system at the current moment based on the origin of the vehicle coordinate system and the x-axis direction of the vehicle coordinate system.

[0023] In one embodiment, the processing module is specifically used to perform optimal estimation of the attitude angle state by using Kalman filtering with the latest attitude angle output by the global navigation satellite system as the observation and the heading angular velocity of the inertial measurement unit as the state recursion input, to obtain the heading angle of the vehicle at the current moment, and to use the heading of the vehicle as the x-axis direction of the vehicle coordinate system.

[0024] In one embodiment, the processing module is specifically used to calculate the comprehensive distance metric between each road shape point and each point in the visual lane line point sequence in the vehicle coordinate system. The comprehensive distance metric is obtained by weighted fusion of Euclidean distance, curvature similarity, and radian angle difference. Using the comprehensive distance metric as the matching cost, the dynamic time warping algorithm is used to solve the correspondence with the minimum global matching cost through dynamic programming.

[0025] In one embodiment, the processing module is specifically used to calculate the optimal rotation matrix from the set of road shape points to the corresponding set of visual lane line point sequences based on the correspondence relationship using singular value decomposition, as a spatial transformation parameter.

[0026] In one embodiment, the processing module is specifically used to remove outliers from the corrected road shape points based on curvature continuity and positional trends; perform linear interpolation on sparse shape points whose spacing is greater than a set spacing threshold and whose curvature of the line connecting adjacent shape points is less than a set curvature threshold; and perform curve fitting based on the road shape points after outlier removal and linear interpolation to obtain the road trajectory.

[0027] In one embodiment, the processing module is specifically used to fit the road shape points after outlier removal and linear interpolation using cubic Bézier curves or cubic spline curves to obtain the road trajectory.

[0028] Thirdly, a vehicle is provided that includes the device as described in any one of the second aspects above.

[0029] In this embodiment, by unifying the road shape points in the navigation map and the visual lane line point sequence into the same vehicle coordinate system, and using the visual lane line point sequence to correct the position of the road shape points in the vehicle coordinate system, and then generating the road trajectory ahead of the vehicle based on the corrected road shape points, it is possible to generate a continuous road trajectory ahead using standard navigation map road shape points without relying on high-precision maps in scenarios where visual lane lines fail. This provides a reference for lateral control, thereby reducing trajectory deviation and swaying phenomena, while adapting to the cost constraints of low computing power platforms. Attached Figure Description

[0030] Figure 1 This is a flowchart illustrating a road trajectory generation method provided in an embodiment of this application.

[0031] Figure 2 This is a schematic diagram of the structure of a road trajectory generation device provided in an embodiment of this application.

[0032] Figure 3 This is a schematic diagram of the architecture of a vehicle provided in an embodiment of this application. Detailed Implementation

[0033] The terminology used in the following embodiments of this application is for the purpose of describing particular embodiments only and is not intended to be limiting of this application. As used in the specification and appended claims of this application, the singular expressions “a,” “an,” “the,” “the,” “the,” and “this” are intended to include the plural expressions as well, unless the context clearly indicates otherwise. It should also be understood that “ / ” means “or,” for example, A / B can mean A or B; “and / or” in the text is merely a description of the relationship between related objects, indicating that three relationships can exist, for example, A and / or B can mean: A alone, A and B simultaneously, and B alone.

[0034] In this application, the reference to "embodiment" means that a specific feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a mutually exclusive, independent, or alternative embodiment. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described in this application can be combined with other embodiments.

[0035] The terms "first" and "second" in the following embodiments of this application are for descriptive purposes only and should not be construed as implying relative importance or implicitly indicating the number of indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature, and in the description of the embodiments of this application, unless otherwise stated, "multiple" means two or more.

[0036] In driver assistance systems, lateral control typically requires obtaining the road trajectory ahead of the vehicle as a reference to ensure stable driving along the desired path. This is especially crucial in complex road structures such as intersections, ramps, and sharp curves, where the continuity and accuracy of the road trajectory ahead are of paramount importance.

[0037] In related technologies, visually perceived lane line information is typically used for lateral control to obtain the trajectory of the road ahead. On regular roads, visual lane lines can provide effective lane-level references. However, in scenarios such as intersections, ramps, and sharp curves, visual lane lines often suffer from missing lines, complex structures, or insufficient effective visible length, making it difficult to obtain continuous and stable road references relying solely on visual lane lines. To compensate for the failure of visual lane lines, some related technologies use high-precision maps; however, the acquisition and maintenance costs of high-precision maps are high, and they are often difficult to deploy directly on low-computing-power platforms.

[0038] Therefore, given the limited reliability of visual lane line detection and the difficulty for low-computing-power platforms to utilize high-precision maps, how to generate continuous and reliable forward road trajectories as a reference for lateral control, thereby reducing vehicle trajectory deviation and "swaying" phenomena, has become a pressing technical problem to be solved in related technologies.

[0039] To address the aforementioned issues, this application embodiment obtains a visual lane line point sequence, unifies the road shape points and the visual lane line point sequence into the vehicle coordinate system, and then uses the visual lane line point sequence to correct the position of the road shape points, making the position of the road shape points closer to the actual geometric shape of the road. As a result, the road trajectory generated based on the corrected road shape points has higher accuracy and can effectively optimize the lateral control performance of the vehicle in scenarios where lane line detection is limited.

[0040] The road trajectory generation method provided in this application can be executed by one or more controllers or processors within a vehicle. In some possible implementations, the method provided in this application can be executed based on a system configured in the vehicle. The specific architecture of the system can be configured according to the actual situation and in conjunction with the method provided in this application. The system can be constructed from hardware, software, or a combination of software and hardware, and no limitations are imposed here.

[0041] The road trajectory generation method provided in this application will be described in detail below with reference to the accompanying drawings.

[0042] Reference Figure 1 This is a flowchart illustrating a road trajectory generation method provided in an embodiment of this application. Figure 1 As shown, the method may specifically include the following S101-S104.

[0043] S101. Obtain the sequence of road shape points and visual lane line points in the navigation map.

[0044] Among them, the visual lane line point sequence is the point sequence on the lane line perceived by the vehicle's vision.

[0045] S102. Unify the road shape points and visual lane line point sequences into the vehicle coordinate system.

[0046] S103. Correct the position of road shape points in the vehicle coordinate system based on the visual lane line point sequence.

[0047] S104. Generate the road trajectory in front of the vehicle based on the corrected road shape points.

[0048] In this embodiment, the navigation map can be, for example, a standard navigation-level map (i.e., a standard definition map, typically used by navigation applications), which differs from a high-definition map. Standard navigation-level maps generally do not contain lane-level detailed information and topological relationships; their road shape points represent the geometric direction of the road in latitude and longitude coordinates. Using a standard navigation-level map instead of a high-definition map allows this solution to remain applicable in scenarios such as national highways and provincial highways not covered by high-definition maps, reducing map acquisition and maintenance costs and expanding the applicable scenarios for the assisted driving system. Road shape points are discrete coordinate points in the navigation map that represent the geometric direction of the road; for example, they can be an ordered sequence of points represented by latitude and longitude. Road shape points provide information about the geometric direction of the road ahead, providing a basic data source for generating road trajectories when visual lane lines are limited. For example, the road shape points obtained in this embodiment can be road shape points corresponding to the road segments matched in the navigation map based on the vehicle's position and heading angle, thus representing the geometric direction of the road ahead.

[0049] The visual lane line point sequence is a sequence of points on the lane lines perceived by the vehicle's vision. This visual lane line point sequence can be, for example, an ordered sequence of points obtained by discretizing continuous lane lines (i.e., visually perceived lane lines) after detecting lane lines using an onboard forward-facing camera. The visual lane line point sequence reflects the real geometric information of the road near the vehicle's current location, and its relative position accuracy is high, serving as a benchmark reference for correcting deviations in the road shape points position of navigation maps. When the reliability of visual lane line detection is limited, its effective length is insufficient, or it is only partially effective, it is impossible to accurately provide a reference for the vehicle's lateral control directly based on the visual lane lines. In such cases, the method provided in this application can be used to introduce road shape points from the navigation map to generate corresponding road trajectories as reference lane lines for the vehicle's lateral control.

[0050] Unifying road shape points and visual lane line point sequences into the vehicle coordinate system places them under the same coordinate reference frame. Unifying two spatial data sources from different origins into a single coordinate system eliminates the incomparability caused by coordinate system differences, providing a necessary prerequisite for establishing the correspondence between the two and performing spatial corrections. Furthermore, using the vehicle coordinate system as a reference ensures that the final road trajectory is a trajectory with the vehicle as the reference frame. This facilitates the vehicle's direct lateral control based on the obtained road trajectory, reducing post-processing steps such as reference frame conversion, improving vehicle lateral control efficiency, and reducing control latency.

[0051] Based on this, the positions of road shape points in the vehicle coordinate system are corrected using a visual lane line point sequence. Specifically, the visual lane line point sequence is used as a positional reference to spatially correct the coordinates of the road shape points. Because the vehicle positioning system is susceptible to the attenuation of Global Navigation Satellite System (GNSS) accuracy and the cumulative errors of the Inertial Measurement Unit (IMU) in non-open environments, the positions of road shape points in the vehicle coordinate system may deviate. By using a visual lane line point sequence with high relative positional accuracy to correct the road shape points, the spatial deviation caused by positioning errors can be effectively eliminated, improving the positional accuracy of road shape points in the vehicle coordinate system.

[0052] Finally, the road trajectory ahead of the vehicle is generated based on the corrected road shape points. This road trajectory can, for example, serve as a reference path for the vehicle's lateral control. The road trajectory generated based on the corrected road shape points more closely matches the geometry of the actual road, providing a more accurate reference path for the lateral control system, thereby effectively preventing vehicle trajectory deviation in scenarios where visual lane line detection is limited, such as intersections and ramps.

[0053] The overall technical approach of the embodiments of this application has been explained above. The implementation path of the correction process will be further described below. In some embodiments, the correction process may include three stages: first, establishing a correspondence between road shape points and visual lane line point sequences in the vehicle coordinate system; second, calculating spatial transformation parameters based on this correspondence; and finally, transforming the positions of the road shape points in the vehicle coordinate system based on the spatial transformation parameters to obtain the corrected road shape points. By decomposing the correction process into establishing a correspondence and spatial transformation, the correction result has global optimality. Compared to point-by-point independent correction, it can better maintain the relative geometric relationships between road shape points and improve the overall spatial consistency of the corrected road shape points.

[0054] Establishing a correspondence between road shape points and the visual lane line point sequence in the vehicle coordinate system, i.e., determining which point in the visual lane line point sequence each road shape point corresponds to. Establishing this correspondence is a prerequisite for subsequent calculation of spatial transformation parameters. By clarifying the one-to-one or many-to-one correspondence between shape points and lane line points, constraints are provided for solving the globally optimal spatial transformation.

[0055] Based on this correspondence, spatial transformation parameters are calculated. These parameters describe the spatial mapping relationship between the set of road shape points and the set of visual lane line point sequences. The spatial transformation parameters express the positional deviation between road shape points and the visual lane line point sequences in a unified mathematical transformation form, ensuring global consistency in the correction of all road shape points and avoiding local discontinuities that may be introduced by point-by-point independent correction.

[0056] The positions of road shape points in the vehicle coordinate system are transformed based on spatial transformation parameters to obtain corrected road shape points. This involves applying spatial transformation parameters to each road shape point to bring its position closer to the actual road position represented by the visual lane lines. By correcting the positions of road shape points through a unified spatial transformation, the corrected road shape points are positioned in the vehicle coordinate system closer to the actual geometric position of the road, while maintaining the relative geometric relationships between the shape points.

[0057] The acquisition method of the visual lane line point sequence mentioned above is described below. In some embodiments, the process of acquiring the visual lane line point sequence may include: firstly, selecting a target lane line from the lane lines perceived by the vehicle's visual perception based on the vehicle's direction of travel; and then performing discrete sampling within a preset range of the target lane line to obtain the visual lane line point sequence. By selecting the lane line most relevant to the current driving intention based on the direction of travel as the correction benchmark, matching deviations caused by using lane lines inconsistent with the direction of travel are avoided; by performing discrete sampling within a preset range, continuous lane lines are transformed into an ordered point sequence, which facilitates subsequent matching calculations with road shape points.

[0058] The vehicle's direction of travel may include, for example, turning left, turning right, or going straight. Lane lines perceived by the vehicle's vision may include, for example, the left lane line, the right lane line, and / or the center line. A target lane line is selected based on the vehicle's direction of travel, ensuring that the selected lane line matches the path the vehicle will be traveling in. This allows subsequent road shape point corrections based on this lane line to more accurately reflect the road geometry information in the vehicle's actual travel direction, improving the targeting and accuracy of the correction.

[0059] Discrete sampling is performed within a preset range of the target lane line to obtain a visual lane line point sequence. The preset range can be, for example, a specific length interval or a specific distance range of the target lane line. By performing discrete sampling within the preset range, the continuous visual lane lines are transformed into a discrete, ordered point sequence, making its data form consistent with road shape points, which facilitates the subsequent establishment of the correspondence between the two using point sequence matching algorithms.

[0060] In an optional embodiment, the target lane line can be discretely sampled within the effective forward-looking range at equal intervals according to longitudinal distance or curve arc length to obtain a preset number of visual lane line points. The preset number can be calibrated according to the visual perception range, processor computing power, and matching accuracy requirements, for example, it can be 25. By discretizing the visual lane line into an ordered sequence of points with a fixed number or approximately fixed density, the stability of matching calculations in different processing cycles can be improved.

[0061] For example, the target lane selection strategy can be as follows for different directions of travel: When the vehicle is turning right, the left lane line in the vehicle's visual perception of the lane lines is selected as the target lane line. When the vehicle turns right, the right lane line usually disappears earlier or is structurally incomplete during the turn, while the left lane line usually has a longer visible range and more stable detection results in right-turn scenarios. Selecting the left lane line as the target lane line provides a more reliable correction benchmark. When the vehicle is turning left, the right lane line in the vehicle's visual perception of the lane lines is selected as the target lane line. When the vehicle turns left, the left lane line usually disappears earlier or is structurally incomplete during the turn, while the right lane line usually has a longer visible range and more stable detection results in left-turn scenarios. Selecting the right lane line as the target lane line provides a more reliable correction benchmark. When the vehicle is traveling straight, the center line of the lane lines in the vehicle's visual perception of the lane lines is selected as the target lane line. When a vehicle is traveling straight, the centerline is most aligned with its trajectory. Choosing the centerline as the target lane line ensures that the correction benchmark closely matches the vehicle's actual path, improving the accuracy of subsequent road shape point correction. By employing the aforementioned strategy of selecting corresponding target lane lines for different directions of travel, the lane line that best matches the vehicle's trajectory can be chosen as the correction benchmark in various scenarios, such as turning and straight-ahead driving. This enhances the adaptability and accuracy of subsequent road shape point correction across diverse driving conditions.

[0062] In an optional embodiment, when the vehicle is traveling in a straight direction, a lane centerline can be generated based on the left and right lane lines perceived by the vehicle's vision. This centerline can then be used as the target lane line based on the target lane line selection strategy described in the previous example. This centerline can be a virtual centerline obtained by equidistant or weighted processing of the left and right lane lines, or it can be a road center reference line determined based on the vehicle's current lane position.

[0063] Of course, it should be noted that the target lane line selection strategy in the above example is only one example. In other possible implementations of this application, the center line can be selected as the target lane line in any direction of vehicle travel, or the corresponding target lane line can be selected by using a selection strategy opposite to the above selection strategy example in the scenario of vehicle turning left and right. There are no restrictions here, and the specific configuration can be made according to the actual implementation needs and situation.

[0064] The acquisition method of the visual lane line point sequence has been explained above. The process of unifying the road shape points and the visual lane line point sequence into the vehicle coordinate system is described below. In some embodiments, since the acquisition times of the road shape points and the visual lane line point sequence may be different, and the vehicle is in continuous motion, it is necessary to unify the two into the vehicle coordinate system at the same moment through time compensation, so as to eliminate the coordinate deviation caused by the difference in acquisition time and vehicle motion, and ensure the spatial consistency of subsequent matching and correction.

[0065] For example, road shape points can be compensated from historical time points to the current vehicle coordinate system based on the vehicle's own speed and heading angular velocity. This historical time point could be, for example, the moment the most recent GNSS data was received. The coordinate transformation of road shape points relies on vehicle position information provided by GNSS, but the GNSS update frequency is typically low, causing the vehicle coordinate system corresponding to the acquired road shape points to lag behind the current time. By using vehicle speed and heading angular velocity for time compensation, the road shape points are extrapolated from the GNSS update time to the current vehicle coordinate system, eliminating the time lag caused by the low frequency of GNSS updates.

[0066] For example, the visual lane line point sequence can be compensated from a historical moment to the current vehicle coordinate system based on the vehicle's own speed and heading angular velocity. This historical moment could be, for example, the moment when the visual perception module outputs the lane line detection results. There is a processing delay between the acquisition time of the visual lane line point sequence and the current moment, during which the vehicle continues to move. By using the vehicle's own speed and heading angular velocity for time compensation, the visual lane line point sequence can be extrapolated from its acquisition time to the current vehicle coordinate system, ensuring that it is within the same coordinate reference frame as the road shape points at that moment.

[0067] In an optional embodiment, when unifying the road shape points and visual lane line point sequences to the vehicle coordinate system, the road shape points can first be transformed from the world coordinate system to the vehicle coordinate system based on the vehicle's position and attitude angle (i.e., heading angle) in the most recently received GNSS data. Then, based on the vehicle's pose changes from a historical moment to the current moment, the road shape points and visual lane line point sequences are compensated to the vehicle coordinate system at the current moment. Road shape points are usually represented in latitude and longitude form, which can be first transformed to local planar coordinates, then translated in combination with the vehicle's position and rotated in combination with the vehicle's heading angle, thereby completing the transformation from the world coordinate system to the vehicle coordinate system. Of course, in some other possible implementations, the world coordinates of the road shape points can be compensated to the current moment based on the vehicle's pose changes from a historical moment to the current moment, and then the compensated road shape points can be transformed from the world coordinate system to the vehicle coordinate system. That is, in this embodiment, the execution order of time compensation and coordinate system transformation of road shape points is not limited.

[0068] The vehicle coordinate system was mentioned above; the construction method of the vehicle coordinate system will be further explained below. In some embodiments, the vehicle coordinate system is constructed by fusing GNSS and IMU information to determine the origin and x-axis direction. GNSS provides absolute position and attitude angles but has a low update frequency, while IMU provides high-frequency velocity and angular velocity data but has accumulated errors. The fusion of the two can achieve continuous, real-time calculation of vehicle position and attitude within the GNSS update interval, meeting the real-time requirements of the assisted driving system.

[0069] For example, the vehicle's speed provided by the IMU can be integrated over time to estimate the vehicle's longitudinal and lateral displacements from the moment the last GNSS data was received to the present moment, and the current vehicle position can be calculated as the origin of the vehicle coordinate system. GNSS update frequencies are typically low (e.g., 1 Hz), while driver assistance systems require higher frequency position information. By integrating the high-frequency speed data (e.g., 50 Hz) provided by the IMU over time, the vehicle position can be continuously calculated between two adjacent GNSS updates, enabling real-time updates of the vehicle coordinate system origin and meeting the real-time requirements of the control system.

[0070] Alternatively, the vehicle's heading angle at the current moment can be obtained by recursively fusing the latest GNSS attitude angle as the initial reference and integrating the IMU's heading angular velocity. This heading (i.e., the vehicle's forward direction) can then be used as the x-axis direction of the vehicle coordinate system. Using the GNSS attitude angle as the initial reference eliminates the absolute error of the IMU's heading angle. At the same time, recursively using the IMU's high-frequency heading angular velocity can obtain a continuous and smooth heading angle estimate within the GNSS update interval. The fusion of these two methods ensures both the absolute accuracy of the heading angle and meets real-time requirements, providing an accurate orientation reference for the vehicle coordinate system.

[0071] Based on the origin and x-axis of the vehicle coordinate system, construct the vehicle coordinate system at the current moment. The x-axis of the vehicle coordinate system represents the vehicle's current direction of travel.

[0072] Of course, in the embodiments of this application, the vehicle coordinate system can also be constructed based on any relevant technology that uses an algorithm to construct a coordinate system centered on the vehicle (or self-moving device) based on GNSS and IMU data, and no limitation is made here.

[0073] In some embodiments, the process of fusing GNSS attitude angles and IMU heading angular velocity to obtain the heading angle at the current moment can be optimally estimated using Kalman filtering. Kalman filtering uses the latest GNSS attitude angle output as the observation and the IMU heading angular velocity as the state recursion input to optimally estimate the attitude angle state, obtaining the vehicle's heading angle at the current moment, with the vehicle's heading as the x-axis direction of the vehicle coordinate system. Kalman filtering can provide a statistically optimal estimate of the attitude angle state, considering GNSS observation noise and IMU recursion cumulative error. Compared to simple recursive superposition, it can obtain a more accurate and smoother heading angle estimation result. Kalman filtering uses the GNSS attitude angle as the observation to periodically correct the IMU recursion result, and simultaneously uses the IMU's high-frequency angular velocity data to perform state prediction between observation updates. This achieves optimal fusion of the characteristics of GNSS (low frequency but high absolute accuracy) and IMU (high frequency but with cumulative drift), resulting in an output heading angle that possesses both absolute accuracy and temporal continuity and smoothness.

[0074] In an optional embodiment, the construction of the vehicle coordinate system can specifically include three stages: position estimation, attitude angle estimation, and coordinate transformation. In the position estimation stage, since the GNSS update frequency is low (e.g., 1 Hz) while the IMU output frequency is high (e.g., 50 Hz), the vehicle position at the moment the latest GNSS data was received can be used as the starting point. The vehicle velocity provided by the IMU is then integrated over time to estimate the longitudinal and lateral displacements of the vehicle from that historical moment to the current moment. In a simplified case, assuming the vehicle mainly moves along the x-axis of the vehicle coordinate system in a short period, the longitudinal displacement of the vehicle at the current moment relative to the moment the latest GNSS data was received is... It can be approximated as: in, The vehicle speed provided by the IMU or chassis The latest time when GNSS positioning data was received. This represents the current moment. When the vehicle is turning, the lateral displacement change can also be estimated by combining the heading angular velocity. Based on the starting position and the aforementioned displacement, the vehicle's current position can be calculated, and this position can be used as the origin of the current vehicle coordinate system.

[0075] During the attitude angle estimation stage, the latest attitude angle output from the GNSS can be used as the initial reference, and the predicted value of the heading angle can be obtained by recursively calculating the high-frequency heading angular velocity from the IMU. This recursion can be understood as: in, This is the latest attitude angle output by GNSS. The heading angular velocity is the output of the IMU. To suppress the accumulated error of the IMU recursion and GNSS observation noise, Kalman filtering can be used to optimally estimate the attitude angle state: using the latest attitude angle output by the GNSS as the observation, and the heading angular velocity of the IMU as the state recursion input, the attitude angle state is predicted and updated to obtain the vehicle's heading angle at the current moment, and the heading angle corresponding to this heading angle is used as the x-axis direction of the current vehicle coordinate system. Of course, in some other possible implementations, the attitude angle state can be optimally estimated without relying on Kalman filtering, and the vehicle's heading angle at the current moment can be directly calculated based on the aforementioned example formula using time integration; this is not a limitation here.

[0076] During the coordinate transformation phase, road shape points can be transformed from the world coordinate system to the current vehicle coordinate system based on the current vehicle position and heading angle. For example, the latitude and longitude of the road shape points can first be transformed to local planar coordinates, then translated according to the vehicle position, and rotated according to the vehicle's heading angle. The transformation relationship can be expressed as: in, This represents the position of the road shape point in world coordinates. This represents the vehicle's position in world coordinates. The rotation matrix is ​​determined by the vehicle's heading angle. This represents the position of the road shape point in the vehicle coordinate system. Through the above position calculation, attitude angle calculation, and coordinate transformation, the vehicle coordinate system at the current moment can be continuously established within the GNSS low-frequency update interval, providing a unified spatiotemporal reference for subsequent road shape point-visual lane line point sequence matching and trajectory generation.

[0077] In an optional embodiment, it can be determined whether new GNSS latitude and longitude information has been received in each processing cycle. If new GNSS latitude and longitude information is received, the vehicle's position and attitude in the world coordinate system are updated based on the new GNSS latitude and longitude information and heading information (attitude angle), and this position and attitude are used as the reference for subsequent coordinate transformations. If no new GNSS latitude and longitude information is received, the position and attitude corresponding to the previously received GNSS latitude and longitude information are used as the reference, and the vehicle's pose change from the previous GNSS time to the current time is recursively calculated using the vehicle speed and heading angular velocity acquired in the current processing cycle. Based on the recursively calculated pose change, the road shape points are compensated to the vehicle coordinate system at the current time. Thus, even in the case of low-frequency GNSS updates or short periods without updates, the time consistency between the road shape points and the current vehicle coordinate system can still be maintained.

[0078] The construction method and coordinate unification process of the vehicle coordinate system have been explained above. Now, returning to the calibration stage, we will explain how to establish the correspondence between road shape points and the visual lane line point sequence. In some embodiments, the correspondence can be established through a combination of distance metrics and Dynamic Time Warping (DTW) algorithms. The distance metrics integrate multiple geometric features, allowing matching to consider not only spatial proximity but also curvature and orientation similarity, thus improving the robustness of the matching. The DTW algorithm solves for the globally optimal match through dynamic programming, avoiding the local optima problem that may result from greedy matching.

[0079] For example, a comprehensive distance metric in the vehicle coordinate system between each road shape point and each point in the visual lane line point sequence can be calculated first. This comprehensive distance metric is obtained by weighted fusion of at least two of Euclidean distance, curvature similarity, and radian angle difference. When matching using only Euclidean distance, incorrect matching may occur if the road shape points shift overall due to positioning errors. By fusing geometric features such as curvature similarity and radian angle difference, even with positional shifts, the correct correspondence can still be identified using the local geometric features of the road (curvature change trend, direction change), significantly improving the robustness and accuracy of matching.

[0080] In an optional embodiment, the aforementioned radian angle difference can be a tangential direction angle difference or a heading angle difference. This difference can be determined based on the direction of the line connecting adjacent points, or it can be determined based on the tangent direction of the locally fitted curve.

[0081] Then, using a comprehensive distance metric as the matching cost, the DTW algorithm can be employed to solve for the correspondence with the minimum global matching cost through dynamic programming. The DTW algorithm searches for the corresponding path with the minimum matching cost globally through dynamic programming, which can handle the differences in sampling density and time scale between road shape points and visual lane line point sequences, avoiding the problem of getting trapped in local optima in point-by-point greedy matching, and ensuring that the globally optimal correspondence is obtained.

[0082] After obtaining the correspondence, spatial transformation parameters can be calculated based on this correspondence. In some embodiments, Singular Value Decomposition (SVD) can be used to calculate the optimal rotation matrix from the road shape point set to the corresponding visual lane line point sequence set, which serves as the spatial transformation parameter. The SVD method can solve for the optimal rigid body transformation between two point sets in the least squares sense, and has the advantages of good numerical stability and high computational efficiency. The resulting rotation matrix can map the road shape point set to the position of the visual lane line point sequence set with minimal overall error. The SVD method solves for the optimal rotation matrix in the least squares sense, minimizing the overall distance error between the transformed road shape point set and the corresponding visual lane line point sequence set, thereby achieving globally optimal correction of the road shape point position and ensuring that the overall spatial distribution of the corrected shape points is as close as possible to the actual road geometry.

[0083] In an optional embodiment, establishing the correspondence between road shape points and the visual lane line point sequence and calculating spatial transformation parameters may specifically include the following steps. First, road shape points can be represented as... The visual lane line point sequence can be represented as : Then, the matching cost is constructed. For the sequence of road shape points and visual lane line points, the comprehensive distance metric between each pair of points is calculated. The result is obtained by weighted fusion of at least two of Euclidean distance, curvature similarity, and radian angle difference (the following example uses weighted fusion of all three), and can be expressed as: in, for Euclidean distance between two points for Local curvature similarity between two points for The difference in the local tangent direction between two points, i.e., the difference in radian angle. , , The preset weights can be determined based on calibration. By weighted fusion of multiple geometric features, the matching cost can better reflect the spatial position, curvature, and directional consistency between road shape points and visual lane line points.

[0084] Then, the correspondence is determined using the dynamic time warping algorithm. Using the comprehensive distance metric as the matching cost, dynamic programming is used to find the corresponding path with the minimum global matching cost. The recursive process of dynamic time warping can be expressed as: in, For the sequence from the start point to the point pair The cumulative minimum matching cost is obtained. Through the above recursion, the optimal correspondence between the road shape point set and the visual lane line point sequence set is obtained. This correspondence can adapt to situations where the two sequences have different lengths and sampling densities, and reduces the probability of mismatches globally.

[0085] Finally, spatial transformation parameters are calculated based on the correspondence. For example, let the matched corresponding point pairs be... ,in For road shape points, These are the corresponding visual lane line points. The first centroid of the road shape point set and the second centroid of the visual lane line point set can be calculated separately, and both sets of points can be decentroided to remove the translation effect. Then, the cross-correlation matrix is ​​constructed: in, and These represent the road shape points and visual lane line points after centroid removal, respectively. The cross-correlation matrix... Perform singular value decomposition: in This involves rotating the set of road shape points to a standard orthogonal basis. That is, scaling is performed on this standard orthogonal basis. The standard orthogonal basis is rotated to the visual lane line point set.

[0086] Based on the decomposition results, the optimal rotation matrix can be obtained: In some possible implementations, after obtaining the optimal rotation matrix, a translation vector can be further calculated based on the first centroid, the second centroid, and the optimal rotation matrix, thereby forming rigid body transformation parameters (i.e., spatial transformation parameters) that include rotation and translation. Based on these spatial transformation parameters, a unified transformation is performed on the road shape points to obtain the corrected road shape points. In this way, the road shape points can be corrected as a whole to a spatial position consistent with the visual lane line point sequence using the optimal spatial transformation in the least squares sense, improving the accuracy and overall consistency of the corrected road shape points.

[0087] Of course, the above is only an example of establishing the correspondence between road shape points and visual lane line point sequences, and solving for the optimal rotation matrix based on the correspondence. In other possible implementations of this application, any algorithm that can be used for point-to-point matching (such as point-by-point greedy matching algorithm, Kalman filter, particle filter or graph optimization algorithm, etc.) can be used to match road shape points and visual lane line point sequences to establish a correspondence. In addition, any transformation equation between road shape points and visual lane line point sequences based on rotation matrix (or introducing translation matrix) can be constructed, and then the optimal solution of the transformation equation of each point pair obtained according to the correspondence can be performed based on any least squares correlation algorithm to obtain the corresponding rotation matrix and translation matrix (if any), which can then be used as spatial transformation parameters.

[0088] The complete implementation path of the correction process has been explained above. The following describes the process of generating road trajectories based on the corrected road shape points. In some embodiments, the generation of road trajectories may include three steps: outlier removal, linear interpolation, and curve fitting. Since the road shape points on navigation maps are sparsely distributed and may contain outliers caused by positioning fluctuations or measurement noise, direct curve fitting may lead to fitting distortion or curve oscillation. By first removing outliers and then interpolating to supplement sparse intervals, a higher-quality and more uniformly distributed data foundation is provided for curve fitting, thereby improving the smoothness and accuracy of the final generated road trajectory.

[0089] In an optional embodiment, before performing outlier removal and interpolation on the road shape points, the road shape points can be filtered based on their longitudinal coordinates in the current vehicle coordinate system, retaining those whose longitudinal coordinates fall within a preset longitudinal range. The preset longitudinal range can be calibrated based on vehicle speed, sensor effective range, and lateral control aiming distance; for example, it could be a range from 10 meters behind the vehicle to 300 meters in front of the vehicle. By removing road shape points that are too far from the current vehicle or clearly located behind it and do not participate in the generation of the forward trajectory, the computational load for subsequent matching and fitting can be reduced, and the influence of irrelevant shape points on the road trajectory generation results can be minimized.

[0090] Outliers are removed from the corrected road shape points based on curvature continuity and positional trends. In road sections with relatively dense shape point distribution (such as curved road sections), lateral positional jumps of individual shape points may occur due to positioning fluctuations or measurement noise. By identifying and removing these outliers based on curvature continuity and positional trends, local oscillations in the fitted curve caused by outliers are avoided, thus improving the smoothness of the fitting results.

[0091] Linear interpolation is performed on sparse points with a spacing greater than a set spacing threshold and a curvature of the line connecting adjacent points less than a set curvature threshold. When the spacing between adjacent points is large and the curvature of the line is small, the road segment can be determined to be close to a straight line. Performing linear interpolation within this sparse interval can enhance the continuity and density uniformity of the point sequence, providing sufficient data support for subsequent curve fitting and avoiding distortion of the fitted curve in this interval due to excessively sparse data points.

[0092] In an optional embodiment, discrete curvature or heading angle change rate can be calculated based on three or more consecutive road shape points. When the lateral deviation, curvature difference, or heading angle difference of the target shape point exceeds the corresponding threshold, the target shape point is identified as an anomaly and removed. Alternatively, when the distance between adjacent shape points is greater than a set distance threshold, and the curvature or heading angle change rate determined by the three adjacent shape points is less than a set curvature threshold, linear interpolation is performed between adjacent shape points according to a set step size.

[0093] The road trajectory is obtained by curve fitting based on the road shape points after outlier removal and linear interpolation. The road shape points after outlier removal and linear interpolation preprocessing are of higher quality and more uniformly distributed. Based on this, curve fitting can generate a more accurate and smoother road trajectory, effectively avoiding the curve distortion or oscillation problems that may occur when directly fitting the original sparse shape points.

[0094] Of course, in other possible implementations of this application, outlier removal and interpolation may not be performed on the corrected road shape points, but curve fitting may be performed directly to obtain the road trajectory; or only outlier removal or only interpolation may be performed and then curve fitting may be performed, which is not limited here.

[0095] In some embodiments, the curve type used for curve fitting can be a cubic Bézier curve or a cubic spline curve. Both of these curves possess good smoothness and local control characteristics, accurately representing continuous geometric changes in the road, and are particularly suitable for describing smooth and continuous changes in vehicle trajectories. Cubic Bézier curves offer good geometric intuitiveness and local control characteristics; the curve shape can be intuitively adjusted through control points, and curve changes are only affected by local control points. Compared to traditional high-order polynomial fitting, they are less prone to global oscillations and can better fit the actual road geometry, making them particularly suitable for describing road sections with curvature changes, such as curves. Alternatively, as another implementation, cubic spline curves can be used to fit the road shape points to obtain the road trajectory. Cubic spline curves ensure the continuity of the second derivative of the curve at each shape point, i.e., curvature continuity, so that the generated road trajectory does not have abrupt curvature changes at the connection points between segments. This provides smooth curvature reference information for vehicle lateral control and helps avoid abrupt changes in control commands. Of course, this application embodiment does not impose specific restrictions on the type of curve used for curve fitting. In this application embodiment, any type of curve (such as spline curve, B-spline, Clothoid curve) or traditional polynomial equation can be used to fit the road shape points. The specific curve type can be configured according to the actual situation or needs.

[0096] In an optional embodiment, when using cubic Bézier curves for fitting, the road shape points can be divided into multiple curve segments according to arc length or curvature changes, and each segment can be fitted with a cubic Bézier curve. The control points of each segment can be determined based on the endpoint position, endpoint tangential direction, and least squares fitting error. Adjacent curve segments satisfy positional continuity and first derivative continuity to ensure smooth connection of the entire road trajectory. When using cubic spline curves for fitting, the road shape points can be used as nodes, and the boundary conditions can be set as natural boundary conditions or tangential boundary conditions consistent with the vehicle heading angle. The curve parameters at each node can be solved, and the position and curvature of the trajectory points can be output according to the spline curve. Through the above piecewise fitting and continuity constraints, the road geometry containing complex curvature changes can be stably represented over a long range.

[0097] In an optional embodiment, after generating the road trajectory ahead of the vehicle, trajectory curvature information can also be calculated based on the road trajectory. The trajectory curvature information may include curvature values, curvature change rates, or heading angle changes corresponding to multiple aiming points on the road trajectory. This allows the road trajectory and trajectory curvature information to be output to lateral control-related units or modules, enabling lateral control to determine lateral deviations based on the road trajectory and generate or correct steering control values ​​based on the trajectory curvature information. Thus, by simultaneously outputting road trajectory and trajectory curvature information, the lateral control smoothness of the vehicle can be improved in scenarios involving ramps, sharp curves, or insufficient lane length.

[0098] Based on the above implementation methods provided in this application, the road shape points and visual lane line point sequences in the navigation map can be unified into the same vehicle coordinate system. The position of the road shape points in the vehicle coordinate system can be corrected using the visual lane line point sequence. Then, the road trajectory in front of the vehicle can be generated based on the corrected road shape points. This allows for the generation of a continuous road trajectory in front of the vehicle using standard navigation map road shape points without relying on high-precision maps in scenarios where visual lane lines fail. This provides a reference for lateral control, thereby reducing trajectory deviation and swaying phenomena, while also adapting to the cost constraints of low-computing-power platforms.

[0099] Based on the examples given above, those skilled in the art will obviously be able to make various equivalent modifications or changes, such as adjusting the relevant logical order, adjusting the relevant parameters, and transforming the relevant formulas. Such modifications or changes also fall within the scope of the embodiments of this application.

[0100] It is understood that, in order to achieve the above-mentioned functions, the device may include hardware structures and / or software modules corresponding to the execution of each function. Those skilled in the art should readily recognize that, based on the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein, the embodiments of this application can be implemented in hardware or a combination of hardware and computer software. Whether a function is executed in a hardware-driven or software-driven manner depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the embodiments of this application.

[0101] This application embodiment can divide functional modules according to the above method example, thereby obtaining an apparatus capable of implementing the above method example. For example, Figure 2 A schematic diagram of a road trajectory generation device provided in an embodiment of this application is shown. Figure 2 As shown, the device may include: an acquisition module 201, used to acquire road shape points and visual lane line point sequences in a navigation map, wherein the visual lane line point sequence is a sequence of points on the lane lines perceived by the vehicle; a processing module 202, used to unify the road shape points and visual lane line point sequences into the vehicle coordinate system; to correct the position of the road shape points in the vehicle coordinate system based on the visual lane line point sequence; and to generate the road trajectory in front of the vehicle based on the corrected road shape points.

[0102] In one embodiment, the processing module 202 is specifically used to establish a correspondence between road shape points and visual lane line point sequences in the vehicle coordinate system; calculate spatial transformation parameters based on the correspondence; and transform the position of the road shape points in the vehicle coordinate system based on the spatial transformation parameters to obtain corrected road shape points.

[0103] In one embodiment, the acquisition module 201 is specifically used to select a target lane line from the lane lines perceived by the vehicle's vision according to the vehicle's direction of travel; and to perform discrete sampling within a preset range of the target lane line to obtain a visual lane line point sequence.

[0104] In one embodiment, the acquisition module 201 is specifically used to select the left lane line of the lane lines perceived by the vehicle as the target lane line when the vehicle's direction of travel is right turn; select the right lane line of the lane lines perceived by the vehicle as the target lane line when the vehicle's direction of travel is left turn; and select the center line of the lane lines perceived by the vehicle as the target lane line when the vehicle's direction of travel is straight.

[0105] In one embodiment, the processing module 202 is specifically used to compensate the road shape points from the historical time to the vehicle coordinate system at the current time based on the vehicle's own speed and heading angular velocity, and to compensate the visual lane line point sequence from the historical time to the vehicle coordinate system at the current time based on the vehicle's own speed and heading angular velocity.

[0106] In one embodiment, the processing module 202 is further configured to perform time integration on the vehicle speed provided by the inertial measurement unit, estimate the longitudinal and lateral displacements of the vehicle from the moment the latest data from the global navigation satellite system was received to the current moment, and calculate the vehicle position at the current moment as the origin of the vehicle coordinate system; using the attitude angle output by the latest global navigation satellite system as the initial reference, and recursively obtaining the heading angle of the vehicle at the current moment by fusing the heading angular velocity of the inertial measurement unit, so as to use the heading of the vehicle as the x-axis direction of the vehicle coordinate system; and construct the vehicle coordinate system at the current moment based on the origin of the vehicle coordinate system and the x-axis direction of the vehicle coordinate system.

[0107] In one embodiment, the processing module 202 is specifically used to perform optimal estimation of the attitude angle state by using Kalman filtering with the latest attitude angle output by the global navigation satellite system as the observation and the heading angular velocity of the inertial measurement unit as the state recursion input, to obtain the heading angle of the vehicle at the current moment and to use the heading of the vehicle as the x-axis direction of the vehicle coordinate system.

[0108] In one embodiment, the processing module 202 is specifically used to calculate the comprehensive distance metric between each road shape point and each point in the visual lane line point sequence in the vehicle coordinate system. The comprehensive distance metric is obtained by weighted fusion of Euclidean distance, curvature similarity and radian angle difference. Using the comprehensive distance metric as the matching cost, the dynamic time warping algorithm is used to solve the correspondence with the minimum global matching cost through dynamic programming.

[0109] In one embodiment, the processing module 202 is specifically used to calculate the optimal rotation matrix from the set of road shape points to the corresponding set of visual lane line point sequences based on the correspondence relationship using singular value decomposition, as a spatial transformation parameter.

[0110] In one embodiment, the processing module 202 is specifically used to remove outliers from the corrected road shape points based on curvature continuity and position trend; perform linear interpolation on sparse shape points whose spacing is greater than a set spacing threshold and whose curvature of the line connecting adjacent shape points is less than a set curvature threshold; and perform curve fitting based on the road shape points after outlier removal and linear interpolation to obtain the road trajectory.

[0111] In one embodiment, the processing module 202 is specifically used to fit the road shape points after outlier removal and linear interpolation using cubic Bézier curves or cubic spline curves to obtain the road trajectory.

[0112] This application also provides a vehicle that includes the aforementioned road trajectory generation device, capable of executing the method described in the foregoing embodiments to generate a road trajectory.

[0113] As an example, Figure 3 A schematic diagram of the architecture of a vehicle according to an embodiment of this application is shown. Figure 3 As shown, the vehicle may include at least: a bus 301, a processor 302, a memory 303, and a communication interface 304. The processor 302, memory 303, and communication interface 304 communicate via the bus 301. The memory 303 stores computer program code, which includes computer instructions. When these computer instructions are executed by the processor 302, the vehicle performs the methods provided in the foregoing embodiments. It should be understood that this application does not limit the number of processors 302 and memories 303 in the vehicle.

[0114] Bus 301 can be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. Bus 301 can be divided into address bus, data bus, control bus, etc. For ease of representation, Figure 3 The bus 301 may be represented by a single line, but this does not mean that there is only one bus or one type of bus. The bus 301 may include a path for transmitting information between various components of an electronic device (e.g., memory 303, processor 302, communication interface 304).

[0115] The processor 302 may include any one or more of the following: central processing unit, microprocessor (MP), digital signal processor (DSP), graphics processor, microcontroller, neural network processing unit, system-on-a-chip, etc.

[0116] Memory 303 may include volatile memory, such as random access memory (RAM). Memory 303 may also include non-volatile memory, such as read-only memory (ROM), flash memory, hard disk drive (HDD), or solid state drive (SSD).

[0117] Communication interface 304 uses a command distribution module, such as, but not limited to, a network interface card or transceiver, to enable communication between the device and other devices or communication networks.

[0118] In the embodiments of this application, the bus 301, processor 302, memory 303 and communication interface 304 described above can be deployed independently, or they can be fully or partially integrated into a system-on-a-chip for deployment, and no restrictions are imposed here.

[0119] Furthermore, in the embodiments of this application, the functional units or modules of the apparatus used to implement the aforementioned method example can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0120] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a readable storage medium. Based on this understanding, the technical solutions of the embodiments of this application, essentially or in other words, the parts that contribute to the prior art, or all or part of the technical solutions, can be embodied in the form of a software product. This software product is stored in a storage medium and includes several instructions to cause a device (which may be a microcontroller, chip, etc.) or processor to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0121] Therefore, embodiments of this application also provide a computer program product, which includes computer program instructions that, when executed by a processor, cause the processor to perform the methods provided in the foregoing embodiments.

[0122] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions within the technical scope disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for generating road trajectories, characterized in that, include: Obtain road shape points and visual lane line point sequences from the navigation map, wherein the visual lane line point sequence is the point sequence on the lane line perceived by the vehicle's vision; Unify the road shape points and the visual lane line point sequence into the vehicle coordinate system; The position of the road shape points in the vehicle coordinate system is corrected based on the visual lane line point sequence; The road trajectory ahead of the vehicle is generated based on the corrected road shape points.

2. The method according to claim 1, characterized in that, The step of correcting the position of the road shape points in the vehicle coordinate system based on the visual lane line point sequence includes: Establish the correspondence between the road shape points and the visual lane line point sequence in the vehicle coordinate system; Calculate the spatial transformation parameters based on the correspondence; The position of the road shape point in the vehicle coordinate system is transformed based on the spatial transformation parameters to obtain the corrected road shape point.

3. The method according to claim 1 or 2, characterized in that, Obtain the visual lane line point sequence, including: The target lane line is selected from the lane lines perceived by the vehicle's vision based on the vehicle's direction of travel. Discrete sampling is performed within a preset range of the target lane line to obtain the visual lane line point sequence.

4. The method according to claim 3, characterized in that, The step of selecting a target lane line from the lane lines perceived by the vehicle's vision based on the vehicle's direction of travel includes: When the vehicle is traveling in the direction of a right turn, the left lane line in the lane lines perceived by the vehicle's vision is selected as the target lane line. When the vehicle is traveling in a left-turn direction, the right lane line in the lane lines perceived by the vehicle's vision is selected as the target lane line. When the vehicle is traveling in a straight direction, the center line of the lane line perceived by the vehicle's vision is selected as the target lane line.

5. The method according to claim 1, characterized in that, The step of unifying the road shape points and the visual lane line point sequence into the vehicle coordinate system includes: The road shape points are compensated from historical time to the vehicle coordinate system at the current time based on the vehicle's own speed and heading angular velocity; The visual lane line point sequence is compensated from historical time to the vehicle coordinate system at the current time based on the vehicle's own speed and heading angular velocity.

6. The method according to claim 1, characterized in that, The method further includes: The vehicle's own speed provided by the inertial measurement unit (IMU) is integrated over time to estimate the longitudinal and lateral displacements of the vehicle from the moment it last received GNSS data to the present moment, and the vehicle's current position is calculated as the origin of the vehicle's coordinate system. Using the latest attitude angle output by GNSS as the initial reference, the heading angle of the vehicle at the current moment is obtained by recursively fusing the heading angular velocity of the IMU, so that the heading of the vehicle is used as the x-axis direction of the vehicle coordinate system; The vehicle coordinate system at the current moment is constructed based on the origin of the vehicle coordinate system and the x-axis direction of the vehicle coordinate system.

7. The method according to claim 6, characterized in that, The method of using the latest attitude angle output by GNSS as the initial reference and recursively obtaining the heading angle of the vehicle at the current moment by fusing the heading angular velocity of the IMU, so as to use the heading of the vehicle as the x-axis direction of the vehicle coordinate system, includes: By using Kalman filtering with the latest GNSS output attitude angle as the observation and the IMU's heading angular velocity as the state recursion input, the attitude angle state is optimally estimated to obtain the vehicle's heading angle at the current moment, and the vehicle's heading is used as the x-axis direction of the vehicle's coordinate system.

8. The method according to claim 2, characterized in that, Establishing the correspondence between the road shape points and the visual lane line point sequence in the vehicle coordinate system includes: Calculate the comprehensive distance metric between each of the road shape points and each point in the visual lane line point sequence in the vehicle coordinate system. The comprehensive distance metric is obtained by weighted fusion of Euclidean distance, curvature similarity and radian angle difference. Using the comprehensive distance metric as the matching cost, the dynamic time warping algorithm is employed to solve for the correspondence with the minimum global matching cost through dynamic programming.

9. The method according to claim 8, characterized in that, The calculation of spatial transformation parameters based on the correspondence includes: Based on the correspondence, the optimal rotation matrix from the set of road shape points to the corresponding set of visual lane line point sequences is calculated using singular value decomposition, and used as the spatial transformation parameter.

10. The method according to claim 1, characterized in that, The process of generating the road trajectory ahead of the vehicle based on the corrected road shape points includes: Based on the curvature continuity and positional trend, outlier points are removed from the corrected road shape points; Linear interpolation is performed on sparse points where the spacing is greater than a set spacing threshold and the curvature of the line connecting adjacent points is less than a set curvature threshold. The road trajectory is obtained by curve fitting based on the road shape points after outlier removal and linear interpolation.

11. The method according to claim 10, characterized in that, The process of obtaining the road trajectory by curve fitting based on road shape points after outlier removal and linear interpolation includes: The road trajectory is obtained by fitting the road shape points after outlier removal and linear interpolation using cubic Bézier curves or cubic spline curves.

12. A road trajectory generation device, characterized in that, include: The acquisition module is used to acquire road shape points and visual lane line point sequences in the navigation map, wherein the visual lane line point sequence is the point sequence on the lane line perceived by the vehicle's vision. The processing module is used to unify the road shape points and the visual lane line point sequence into the vehicle coordinate system; correct the position of the road shape points in the vehicle coordinate system based on the visual lane line point sequence; and generate the road trajectory in front of the vehicle based on the corrected road shape points.

13. A vehicle, characterized in that, Includes the road trajectory generation device as described in claim 12.

Images