Vehicle control methods and storage media
By combining vehicle sensor data and high-precision maps, lane position errors are evaluated in real time, and confidence scores are assessed using a Bayesian model. This solves the safety problem of autonomous driving caused by the lag in high-precision map updates and enables safe and reliable path planning for vehicles in complex scenarios.
Patent Information
- Application Number
- CN202510592896.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-05-08
- Publication Date
- 2026-11-10
AI Technical Summary
The inability to update existing high-precision maps in a timely manner causes autonomous vehicles to be unable to accurately identify road conditions in complex scenarios such as urban construction, affecting driving safety and reliability.
By combining vehicle sensor perception data and high-precision maps, lane position errors are evaluated in real time. A perception map is generated by fusing LiDAR and camera data, and a confidence score is evaluated using a Bayesian model, enabling flexible adjustment of vehicle paths.
It improves the safety and flexibility of path planning for autonomous vehicles in complex scenarios such as construction, ensuring that vehicles can adjust navigation strategies in a timely manner and avoid driving risks caused by road changes.
Smart Images

Figure CN123061401A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of intelligent driving technology, and in particular to a vehicle control method and storage medium. Background Technology
[0002] In the field of autonomous driving, vehicle routes are typically planned based on high-precision maps. These high-precision maps are created by scanning roads with pre-set scanning devices or surveying vehicles to collect environmental data and the spatial location of lanes.
[0003] However, with the accelerating pace of urbanization, urban roads are frequently under construction or maintenance, and may also contain fences, roadblocks, temporarily parked vehicles, or other obstacles, causing changes in lane information. In these situations, relying solely on historical lane information stored in high-precision maps for route planning will significantly impact the safety and reliability of autonomous driving. Summary of the Invention
[0004] One objective of this application is to provide a vehicle control method and a computer-readable storage medium to improve the safety and flexibility of autonomous driving.
[0005] In a first aspect, embodiments of this application provide a vehicle control method, comprising: obtaining the spatial position of a first lane centerline corresponding to each perceived lane based on perception data collected by sensors on the vehicle, wherein one perceived lane corresponds to a real lane in a driving road; acquiring a high-precision map and obtaining the spatial position of a second lane centerline corresponding to each perceived lane based on the high-precision map; obtaining a position error corresponding to each perceived lane based on the spatial positions of the first and second lane centerlines corresponding to each perceived lane; scoring the position error corresponding to each perceived lane to obtain a confidence score for each perceived lane; and controlling the vehicle to perform path planning based on the confidence score.
[0006] This application embodiment uses a confidence score to predict whether a lane change will occur in advance, enabling the vehicle to plan a safe, reliable, and efficient route in a timely manner. Furthermore, this application embodiment scores the positional error of each perceived lane to obtain a confidence score. The confidence score clearly expresses the probability of a lane change, providing the vehicle with accurate and reliable decision-making information and helping the vehicle to proactively make more reliable and safer navigation strategies.
[0007] In some embodiments, the sensors on the vehicle include a lidar and a camera. Obtaining the spatial position of the first lane centerline corresponding to each sensing lane based on the perception data collected by the sensors on the vehicle includes: fusing point cloud data collected by the lidar and image data collected by the camera to obtain a perception map corresponding to the driving road; and inputting the perception map into a pre-trained road structure perception model to obtain the spatial position of the first lane centerline corresponding to each sensing lane. By fusing point cloud data and image data, an accurate perception map can still be obtained even in poor lighting conditions, rainy conditions, or foggy conditions where visibility is unclear, thus reliably ensuring the extraction of the spatial position of the first lane centerline corresponding to each sensing lane from the perception map.
[0008] In some embodiments, the high-precision map includes the spatial location of at least one standard lane, and one standard lane corresponds to one real lane; obtaining the spatial location of the second lane centerline corresponding to each perceived lane based on the high-precision map includes: acquiring the current pose of the vehicle; searching for the standard lane corresponding to each perceived lane in the high-precision map based on the current pose of the vehicle and preset filtering conditions; and using the spatial location of the lane centerline of the standard lane corresponding to each perceived lane as the spatial location of the second lane centerline corresponding to each perceived lane.
[0009] Based on the current position of the vehicle, this application embodiment filters a few or highly correlated standard lanes in the high-precision map, thereby reducing the amount of calculation on the center line of the second lane and improving the speed of lane change judgment.
[0010] In some embodiments, the driving road includes at least one real lane, the at least one real lane includes a first real lane, and the sensing lane corresponding to the first real lane includes the first sensing lane; obtaining the position error corresponding to each sensing lane based on the spatial position of the center line of the first lane corresponding to each sensing lane and the spatial position of the center line of the second lane corresponding to each sensing lane includes: mapping the spatial position of the center line of the first lane corresponding to the first sensing lane to the world coordinate system to obtain the spatial position of the center line of the third lane corresponding to the first sensing lane; mapping the spatial position of the center line of the second lane corresponding to the first sensing lane to the world coordinate system to obtain the spatial position of the center line of the fourth lane corresponding to the first sensing lane; obtaining at least one error evaluation index based on the spatial position of the center line of the third lane corresponding to the first sensing lane and the spatial position of the center line of the fourth lane corresponding to the first sensing lane; and obtaining the position error corresponding to the first sensing lane based on at least one error evaluation index.
[0011] In some embodiments, at least one of the error evaluation indicators includes a first type of evaluation indicator. Obtaining at least one error evaluation indicator based on the spatial positions of the center lines of the third and fourth lanes corresponding to the first sensing lane includes: performing discrete point sampling on the center line of the third lane corresponding to the first sensing lane to obtain multiple current position points; performing discrete point sampling on the center line of the fourth lane corresponding to the first sensing lane to obtain multiple standard position points, wherein one standard position point corresponds to one current position point; calculating a first distance value between each current position point and each corresponding standard position point; and generating the first type of evaluation indicator based on the multiple first distance values, wherein the first type of evaluation indicator is used to evaluate the distance error between the center line of the third lane corresponding to the first sensing lane and the center line of the fourth lane corresponding to the first sensing lane.
[0012] In some embodiments, the error evaluation index further includes a second type of evaluation index, and the method further includes: calculating multiple angle differences, wherein the angle difference is the absolute value of the difference between the first tangent angle of the current position point and the second tangent angle of the standard position point corresponding to the current position point; generating the second type of evaluation index based on the multiple angle differences, the second type of evaluation index being used to evaluate the angle error between the center line of the third lane corresponding to the first sensing lane and the center line of the fourth lane corresponding to the first sensing lane.
[0013] In some embodiments, the error evaluation index further includes a third type of evaluation index, and the method further includes: calculating a plurality of curvature difference values, wherein the curvature difference value is the absolute value of the difference between the first curvature of the current position point and the second curvature of the standard position point corresponding to the current position point; generating the third type of evaluation index based on the plurality of curvature difference values, the third type of evaluation index being used to evaluate the curvature error between the center line of the third lane corresponding to the first sensing lane and the center line of the fourth lane corresponding to the first sensing lane.
[0014] This application proposes using the difference in distance, tangential angle, and curvature between the center lines of the third and fourth lanes to characterize the positional error between the center lines of the first and second lanes. The differences in distance, tangential angle, and curvature cover the shape change characteristics of various lines. When a lane is blocked by an obstacle, causing any change in the shape of the center line of the first lane, the differences in distance, tangential angle, and curvature can reliably and accurately describe the changed center line of the first lane. Therefore, this application proposes that by using three evaluation indicators—the differences in distance, tangential angle, and curvature—the positional error between the center lines of the first and second lanes can be accurately and reliably evaluated.
[0015] In some embodiments, the driving road includes at least one real lane, the at least one real lane includes a first real lane, and the sensing lane corresponding to the first real lane includes the first sensing lane; the step of scoring based on the position error corresponding to each sensing lane to obtain a confidence score for each sensing lane includes: obtaining a first prior probability and a second prior probability, the first prior probability being the prior probability when the real lane corresponding to the first sensing lane changes, and the second prior probability being the prior probability when the real lane corresponding to the first sensing lane does not change; inputting the position error corresponding to the first sensing lane into a preset first probability distribution model to obtain a first observation probability when the real lane corresponding to the first sensing lane changes; inputting the position error corresponding to the first sensing lane into a preset second probability distribution model to obtain a second observation probability when the real lane corresponding to the first sensing lane does not change; inputting the first prior probability, the second prior probability, the first observation probability, and the second observation probability into a preset Bayesian model to obtain a confidence score for the first sensing lane. Both the first and second prior probabilities are based on historical data of error distribution or the experience of domain experts. By fusing these two prior probabilities and the two observation probabilities through a Bayesian model, the probability of position error occurring when a lane change occurs can be comprehensively assessed, which is more accurate and reliable than relying solely on the current observation probability.
[0016] In some embodiments, controlling the vehicle to perform path planning based on the confidence score includes: when the confidence score of the perceived lane corresponding to the actual lane where the vehicle is located is greater than or equal to a preset score threshold, adjusting the vehicle's navigation trajectory to obtain an updated navigation trajectory.
[0017] Secondly, embodiments of this application provide a computer-readable storage medium storing a computer program, the computer program including program instructions, which, when executed by a processor, cause the processor to perform the aforementioned vehicle control method. Attached Figure Description
[0018] To more clearly illustrate the technical solutions of the embodiments of this application, the drawings used in the description of the embodiments of this application will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0019] Figure 1 A schematic diagram of the circuit structure of a vehicle provided in an embodiment of this application;
[0020] Figure 2 A schematic diagram of a road scene provided for an embodiment of this application;
[0021] Figure 3 A schematic diagram of a road scene provided for an embodiment of this application;
[0022] Figure 4 A schematic flowchart of a vehicle control method provided in an embodiment of this application;
[0023] Figure 5a A schematic diagram of a straight lane provided for an embodiment of this application;
[0024] Figure 5b A schematic diagram of a curved lane provided for an embodiment of this application;
[0025] Figure 6a A schematic diagram of a straight lane provided for an embodiment of this application;
[0026] Figure 6b A schematic diagram of a straight lane provided for an embodiment of this application;
[0027] Figure 7 A schematic diagram illustrating a coordinate system transformation method provided in an embodiment of this application;
[0028] Figure 8 A schematic diagram of a road scene provided for an embodiment of this application;
[0029] Figure 9 A schematic diagram of a vehicle control device provided in an embodiment of this application;
[0030] Figure 10 This is a schematic diagram of a controller provided in an embodiment of this application. Detailed Implementation
[0031] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application. All other embodiments obtained by those skilled in the art based on the embodiments in this application without inventive effort are within the scope of protection of this application.
[0032] It should be noted that, unless there is a conflict, the various features in the embodiments of this application can be combined with each other, all of which are within the protection scope of this application. Furthermore, although functional modules are divided in the device schematic diagram and a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in a different order than the module division in the device or the order in the flowchart. Moreover, the terms "first," "second," and "third" used in this application do not limit the data or execution order, but only distinguish identical or similar items with essentially the same function and effect.
[0033] Due to projects such as urban rail transit construction, municipal infrastructure renovation, and building construction, construction barriers frequently appear on existing roadways. These barriers alter the original road layout, causing changes in lane information, such as dashed lines becoming solid lines or barriers preventing vehicles from passing. Existing high-precision map update mechanisms suffer from high data collection costs and long update cycles, making it impossible to record changes in urban roads in a timely manner.
[0034] For autonomous vehicles, existing autonomous driving technologies heavily rely on high-precision map data for path planning and driving decisions. In complex construction scenarios, due to outdated map information, autonomous vehicles cannot accurately identify road conditions and make incorrect driving decisions, posing a serious challenge to driving safety.
[0035] This application embodiment combines perception data and high-precision maps to score each lane in real time and detect whether the actual lane has changed in real time. This enables the vehicle to plan its path quickly, safely, and efficiently, effectively improving the safety and flexibility of autonomous driving.
[0036] In one embodiment, the vehicle includes a gasoline-powered vehicle, an electric vehicle, and a hybrid vehicle, and the vehicle may support autonomous driving functions or driver assistance functions. See also... Figure 1 The vehicle 100 includes a lidar 11, a camera 12, a wheel speed sensor 13, an inertial measurement unit (IMU) 14, a positioning sensor 15, and a controller 16. The controller 16 is communicatively connected to the lidar 11 and the camera 12.
[0037] The lidar 11 is mounted on top of or around the vehicle 100. The lidar 11's transmitting module emits a laser beam towards the target object. The echo beam formed by the laser beam's reflection from the target object is received by the lidar's receiving module, forming an echo signal. The lidar 11's processor obtains point cloud data based on the echo signal, thereby acquiring three-dimensional spatial information about the target object and its surrounding environment. In some embodiments, the lidar 11 is one of a mechanical lidar, an optical phased array (OPA) solid-state lidar, a microelectromechanical system (MEMS) solid-state lidar, or a flash solid-state lidar.
[0038] Camera 12 is used to acquire image data, including monocular, binocular, or multi-view cameras. Wheel speed sensor 13 is mounted on the wheel hub of vehicle 100 to acquire wheel speed data. Inertial measurement unit 14 is mounted on vehicle 100. Inertial measurement unit 14 includes an accelerometer and a gyroscope; the accelerometer is used to measure the vehicle's acceleration in three axes, and the gyroscope is used to measure the vehicle's angular velocity. Positioning sensor 15 is mounted on vehicle 100 to acquire real-time dynamic position data based on the Real-Time Kinematic method. Positioning sensor 15 includes a BeiDou navigation system-based locator or a GPS locator. Controller 16 is used to control the vehicle for path planning.
[0039] In one embodiment, to illustrate the vehicle control method in detail, this application embodiment combines... Figure 2 and Figure 3 A general explanation is provided below: Please refer to [link / reference]. Figure 2 The driving route of vehicle 100 includes multiple real lanes, such as the first real lane 21, the second real lane 22, and the third real lane 23. Traffic signs for the first real lane 21 include straight ahead and left turn. The traffic sign for the second real lane 22 is straight ahead, and the traffic sign for the third real lane 23 is right turn. Vehicle 100 is in the second real lane 22, and the driver needs to drive straight through the intersection in the second real lane 22. The vehicle 100's storage module stores a high-definition map or a high-definition map of the local environment surrounding the vehicle, which is obtained in real time from a cloud server. The high-definition map contains historical map information of the driving route of vehicle 100. Alternatively, the high-definition map contains spatial information of each standard lane when there are no obstacles such as construction barriers in the driving route; one standard lane corresponds to one real lane.
[0040] like Figure 2As shown, when there are no construction barriers obstructing the road where vehicle 100 is located, the lane center line 21a of the first real lane 21 can fully and completely represent the lane shape of the first real lane 21. The lane center line 22a of the second real lane 22 and the lane center line 23a of the third real lane 23 can also fully and completely represent the lane shapes of the second real lane 22 and the third real lane 23.
[0041] Please see Figure 3 When a construction fence 24 appears in the second real lane 22, the vehicle 100 generates a perception map (such as a bird's-eye view) based on the point cloud data collected by the lidar 11 and the image data collected by the camera 12. Then, based on the perception map, it obtains the perception lane for each real lane within the perception map, as well as the semantic information of each real lane (such as straight-ahead signs, left-turn signs, right-turn signs, and zebra crossings). The number of perception lanes corresponds to the number of real lanes in the driving road; one perception lane corresponds to one real lane. A perception lane is a passable area composed of multiple discrete points obtained based on the perception map. By obtaining the lane centerline of the perception lane corresponding to each real lane (such as lane centerline 21b, lane centerline 22b, and lane centerline 23b), the passable area within the driving road can be perceived in real time, thereby improving the accuracy of vehicle path planning.
[0042] Since the first real lane 21 and the third real lane 23 are not affected by the construction fence, the lane shape of the first real lane 21 (which defines the passable area of the lane) and the lane shape of the third real lane 23 are not affected. The lane centerline 21b approximately coincides with the lane centerline 21a, and the lane centerline 23b approximately coincides with the lane centerline 23a. Figure 3 In order to clearly display lane center lines 21b and 21a, and lane center lines 23b and 23a, lane center lines 21b and 21a are displayed separately, and lane center lines 23b and 23a are displayed separately, but in reality they are approximately overlapping.
[0043] Because the second real lane 22 is affected by the construction fence 24, the lane shape of the second real lane 22 changes, and a large positional error occurs between the lane centerline 22b and the lane centerline 22a, indicating that the second real lane 22 has changed. If the vehicle 100 still needs to continue straight, since the third real lane 23 is a right-turn lane, the vehicle 100 can switch to the first real lane 21 and then continue straight. Based on the above vehicle control method, the vehicle 100 can reliably detect whether each lane has changed in advance. When the lane in which the vehicle 100 is located changes, it can promptly and flexibly adjust the navigation path and switch to a drivable lane to continue driving. This improves driving safety and driving flexibility, and avoids the driving trouble caused by entering impassable areas.
[0044] In one embodiment, this application provides a vehicle control method. Please refer to... Figure 4 The vehicle control methods include S41 to S45.
[0045] S41, based on the perception data collected by the sensors on the vehicle, obtain the spatial position of the first lane centerline corresponding to each perception lane.
[0046] In one embodiment, step S41 includes: performing data fusion based on point cloud data collected by lidar and image data collected by camera to obtain a perception map corresponding to the driving road; inputting the perception map into a pre-trained road structure perception model to obtain the spatial position of the first lane centerline corresponding to each perception lane.
[0047] The roadway in which the vehicle travels has one or more real lanes, and one perceived lane corresponds to one real lane. The perceived map is a bird's-eye view (BEV) used to comprehensively and completely present the lane information of the roadway. The lane boundaries of the perceived lanes are fitted by multiple discrete points, and the road structure perception model (deep learning model, such as the BEVFusion model or SuperFusion model) is used to extract the lane centerline of the perceived lane as the first lane centerline. The first lane centerline reflects the shape of the perceived lane within the perceived map. For example, based on the DAB-DETR model, a large amount of map sample data labeled with lane centerlines is input into the DAB-DETR model for training, thereby obtaining the road structure perception model.
[0048] In one embodiment, obtaining the spatial position of the first lane centerline corresponding to each perceived lane based on perception data collected by sensors on the vehicle includes: fusing point cloud data collected by LiDAR and image data collected by cameras to obtain a perception map; extracting the perceived lane lines from the perception map, which include left and right lane lines; and determining the spatial position of the centerline located between the left and right lane lines as the spatial position of the first lane centerline based on the spatial positions of the left and right lane lines.
[0049] In some embodiments, the centerline of the first lane may be a straight line or a curve. See also Figure 5a The sensing lane 51 is a straight lane. The solid boundary of the sensing lane 51 is fitted from multiple discrete points, and the dashed line is the centerline of the first lane extracted using the road structure sensing model. In other embodiments, please refer to... Figure 5b The sensing lane 53 is a curved lane, and the center line 54 of the first lane of the sensing lane 53 is a curve.
[0050] Due to the influence of road obstacles on the sensing lane, the shape of the sensing lane changes, and the center line of the first lane of the sensing lane also changes. Please refer to [link / reference]. Figure 6a When a square obstacle 55 exists in the actual lane, the sensing lane 51 is no longer a straight lane; the center line of the first lane of the sensing lane 51 becomes a broken line 56. (See also...) Figure 6b When there is a circular obstacle 57 in the actual lane, the center line of the first lane of the sensing lane 51 is a curve 58.
[0051] Step S42: Obtain a high-precision map and, based on the high-precision map, obtain the spatial position of the center line of the second lane corresponding to each sensing lane.
[0052] High-definition maps (lanemaps) are typically obtained by scanning driving roads manually or with data collection equipment, corresponding to historical information about those roads. High-definition maps usually include road geometry information, lane information, and traffic sign and marking information. Road geometry information accurately describes the shape, length, width, radius of curvature, slope, and elevation changes of roads, providing vehicles with accurate driving path references. Lane information records in detail the number of lanes, their width, lane line types (solid lines, dashed lines, etc.), lane boundary positions, and connections between lanes, assisting vehicles in lane keeping, lane changing, and other maneuvers. Traffic sign and marking information includes the location, shape, content, and meaning of various traffic signs (such as speed limit signs, no-entry signs, and directional signs) and markings (such as straight-ahead signs, left-turn signs, pedestrian crossings, and stop lines), enabling vehicles to recognize and comply with traffic rules.
[0053] High-precision maps include the location information of multiple standard lanes, with one standard lane corresponding to one real-world lane. When a real-world lane changes, a significant positional error will occur between the standard lane corresponding to that real-world lane and the perceived lane corresponding to that real-world lane. By using the positional deviation between the lane centerlines as a comparison benchmark, the change in the real-world lane can be determined.
[0054] In one embodiment, step S42 includes: obtaining the current pose of the vehicle; finding the standard lane corresponding to each sensing lane in the high-precision map according to the current pose of the vehicle and preset filtering conditions; and using the spatial position of the lane centerline of the standard lane corresponding to each sensing lane as the spatial position of the second lane centerline corresponding to each sensing lane.
[0055] The current pose includes the vehicle's position, attitude, or orientation on the road. In some embodiments, the vehicle's current pose data is determined directly based on data collected by a single sensor, for example, based on real-time dynamic position data. In other embodiments, the vehicle's current pose is obtained by fusing point cloud data, wheel speed data, inertial navigation data, and real-time dynamic position data based on a preset Kalman filter. The Kalman filter can be an Error-State Kalman Filter (ESKF). The system state variables of the ESKF include pose state variables, wheel speed state variables, inertial navigation state variables, and real-time dynamic position state variables. The filtering process of the ESKF includes a prediction process and a correction process.
[0056] During prediction, the error state Kalman filter integrates the inertial navigation data, outputs the predicted pose according to the frequency of the inertial measurement unit, and updates the error state covariance matrix. During correction, the error state Kalman filter calculates the residuals between the observed measurements and the system state variables, and calculates the Kalman gain based on the measurement covariance matrix and the system state covariance matrix. The Kalman gain is then weighted with the residuals of each sensor and added to the predicted pose to obtain the vehicle's current pose data.
[0057] The residual between the observed measurement and the system state variable is calculated as follows: (1) When the observed measurement is the observed wheel speed, the residual between the observed wheel speed and the wheel speed state variable is calculated as follows: the observed wheel speed is processed by coordinate transformation to obtain the transformed wheel speed, and the transformed wheel speed is subtracted from the wheel speed state variable to obtain the wheel speed residual. (2) When the observed measurement is the observed point cloud, the observed point cloud is matched with the pre-built point cloud map based on the ICP algorithm to obtain the matched pose, and the matched pose is subtracted from the pose state variable to obtain the point cloud residual. (3) When the observed measurement is the observed real-time dynamic position, the observed real-time dynamic position is subtracted from the real-time dynamic position state variable to obtain the position residual. The embodiments of this application fuse data from multiple types of sensors through a Kalman filter, which can accurately obtain the current pose of the vehicle so as to accurately and reliably find the spatial position of the center line of the second lane.
[0058] In one embodiment, finding the standard lane corresponding to each sensing lane in a high-precision map based on the vehicle's current pose and preset filtering conditions includes: mapping the vehicle's current pose to the coordinate system of the high-precision map to obtain the converted current pose; and finding the standard lane that meets the preset filtering conditions in the high-precision map based on the converted current pose.
[0059] Vehicles face complex driving environments during operation, such as sudden braking, sudden stops, or long skidding distances. To obtain more accurate and reliable pose data, this application's embodiments correct the converted current pose. In one embodiment, real-time dynamic position data is acquired and converted to the coordinate system of a high-precision map to obtain converted real-time dynamic position data. Based on the converted real-time dynamic position data, the converted current pose is corrected to obtain the corrected current pose. Based on the corrected current pose, standard lanes that meet preset filtering conditions are searched among all standard lanes in the high-precision map.
[0060] In some embodiments, correcting the converted pose data based on the converted real-time dynamic position data to obtain the corrected current pose includes: calculating the average value between the converted real-time dynamic position data and the converted current pose to obtain the corrected current pose.
[0061] In some embodiments, correcting the converted current pose based on the converted real-time dynamic position data to obtain the corrected current pose includes: weighting the converted real-time dynamic position data and the converted current pose to obtain the corrected current pose.
[0062] This application embodiment uses real-time dynamic position data to correct the current pose, making up for the lack of reliability of the error state Kalman filter in the face of various extreme abnormal conditions, and enhancing the vehicle's ability to obtain accurate and reliable pose data in harsh environments.
[0063] In some embodiments, the preset filtering condition refers to: calculating the point-to-line distance from the current pose of the corrected vehicle to the center line of each standard lane; retaining all standard lanes whose point-to-line distance is less than a preset distance threshold, the preset distance threshold being related to the width of the actual lanes of the driving road.
[0064] Based on the current pose data, this application embodiment selects a few or highly correlated standard lanes in the high-precision map, thereby reducing the amount of calculation about the center line of the second lane and improving the speed of lane change judgment.
[0065] S43. Based on the spatial position of the center line of the first lane corresponding to each sensing lane and the spatial position of the center line of the second lane corresponding to each sensing lane, the position error corresponding to each sensing lane is obtained.
[0066] Position error is the difference between the center line of the first lane and the center line of the corresponding second lane. Position error is used to represent the difference between current road conditions and historical road conditions. The larger the position error, the higher the probability that a lane change will occur.
[0067] In some embodiments, the driving road includes at least one real lane, the at least one real lane includes a first real lane, and the sensing lane corresponding to the first real lane includes a first sensing lane. Taking the first sensing lane as an example, step S43 includes: mapping the spatial position of the center line of the first lane corresponding to the first sensing lane to the world coordinate system to obtain the spatial position of the center line of the third lane corresponding to the first sensing lane; mapping the spatial position of the center line of the second lane corresponding to the first sensing lane to the world coordinate system to obtain the spatial position of the center line of the fourth lane corresponding to the first sensing lane; obtaining at least one error evaluation index based on the spatial positions of the center lines of the third and fourth lanes corresponding to the first sensing lane; and obtaining the position error corresponding to the first sensing lane based on the at least one error evaluation index.
[0068] Please see Figure 7The perception map is configured with a perception coordinate system X1Y1Z1. The transformation matrix between the perception coordinate system X1Y1Z1 and the world coordinate system X0Y0Z0 includes a first rotation matrix P1 and a first translation matrix T1. Based on the first rotation matrix P1 and the first translation matrix T1, the spatial position of the center line R1 of the first perception lane is mapped to the world coordinate system to obtain the spatial position of the center line R3 of the third lane corresponding to the first perception lane. The high-precision map is configured with a global coordinate system X2Y2Z2. The transformation matrix between the global coordinate system X2Y2Z2 and the world coordinate system X0Y0Z0 includes a second rotation matrix P2 and a second translation matrix T2. Based on the second rotation matrix P2 and the second translation matrix T2, the spatial position of the center line R2 of the second lane corresponding to the first perception lane is mapped to the world coordinate system to obtain the spatial position of the center line R4 of the fourth lane corresponding to the first perception lane.
[0069] Error evaluation metrics are used to evaluate the positional error between the center lines of the third and fourth lanes. In one embodiment, the error evaluation metrics include a first type of evaluation metric, which is used to evaluate the distance error between the center lines of the third and fourth lanes corresponding to the first sensing lane. Based on the spatial positions of the center lines of the third and fourth lanes corresponding to the first sensing lane, at least one error evaluation metric is obtained, including: discretely sampling the center lines of the third lane to obtain multiple current position points (N1, N2, N3…, Ni), where Ni is the i-th current position point. The center line of the fourth lane includes multiple standard position points (M1, M2, M3…, Mi), where Mi is the i-th standard position point, and the standard position point Mi corresponds to the current position point Ni. In one embodiment, a world coordinate system is established with any point as the origin, the lane orientation as the Z-axis, and the direction perpendicular to the ground as the X-axis. The Z-axis coordinates of the standard position point Mi and the current position point Ni are the same or approximately the same. Calculate the first distance value between each current location point and its corresponding standard location point; generate a first-class evaluation index based on multiple first distance values. For example, the average of multiple first distance values can be used as the first-class evaluation index.
[0070] In one embodiment, the error evaluation index further includes a second type of evaluation index. At least one error evaluation index is obtained based on the spatial positions of the center lines of the third and fourth lanes corresponding to the first sensing lane. The index further includes: calculating multiple angle differences, where each angle difference is the absolute value of the difference between the first tangent angle of the current position point on the center line of the third lane and the second tangent angle of the standard position point corresponding to the current position point on the center line of the fourth lane; and generating a second type of evaluation index based on the multiple angle differences. This second type of evaluation index is used to evaluate the angular error between the center lines of the third and fourth lanes corresponding to the first sensing lane. In some embodiments, generating the second type of evaluation index based on multiple angle differences includes: calculating the average of the multiple angle differences to obtain the second type of evaluation index. Alternatively, the mode is selected from the multiple angle differences as the second type of evaluation index.
[0071] In some embodiments, the error evaluation index includes a third type of evaluation index. At least one error evaluation index is obtained based on the spatial positions of the center lines of the third and fourth lanes corresponding to the first sensing lane. The index further includes: calculating multiple curvature difference values, where the curvature difference is the absolute value of the difference between the first curvature of the current position point on the center line of the third lane and the second curvature of the standard position point corresponding to the current position point on the center line of the third lane; and generating a third type of evaluation index based on the multiple curvature difference values. This third type of evaluation index is used to evaluate the curvature error between the center lines of the third and fourth lanes corresponding to the first sensing lane. In some embodiments, generating the third type of evaluation index based on multiple curvature difference values includes: calculating the average value of the multiple curvature difference values to obtain the third type of evaluation index. In other embodiments, generating the third type of evaluation index based on multiple curvature difference values includes: selecting the mode from the multiple curvature difference values as the third type of evaluation index.
[0072] In one embodiment, obtaining the position error corresponding to the first sensing lane based on at least one error evaluation index includes: when the error evaluation index includes only one type of evaluation index, directly using the error evaluation index as the position error. In other embodiments, when the error evaluation index includes at least two types of error evaluation indexes, weighting the at least two error evaluation indexes to obtain the position error.
[0073] The location error is obtained by weighting at least two error evaluation indicators, including: normalizing the first, second, and third evaluation indicators respectively to obtain the first, second, and third normalized values; multiplying the first normalized value by the first weight coefficient to obtain the first weighted value; multiplying the second normalized value by the second weight coefficient to obtain the second weighted value; multiplying the third normalized value by the third weight coefficient to obtain the third weighted value; and summing the first, second, and third weighted values to obtain the location error, wherein the sum of the first, second, and third weight coefficients is equal to the natural number 1.
[0074] The three evaluation indicators—distance error, angle error, and curvature error—cover the shape change characteristics of various lines. Therefore, the embodiments of this application perform weighted processing based on one or more of the first, second, and third evaluation indicators to reliably and accurately describe the positional error between the center lines of the first and second lanes, thereby reliably and accurately determining whether the actual lane has changed.
[0075] In some embodiments, to enhance robustness, the lane centerline information at consecutive time points is tracked, and the stability of the position error over time is evaluated to determine whether there is a sudden change. For example, the absolute value of the difference between two position errors at adjacent time points is calculated to obtain a secondary error value. When the secondary error value is greater than or equal to a preset difference threshold, the first lane centerline is processed based on a preset Kalman filter to obtain a smoothed first lane centerline.
[0076] In some embodiments, the quadratic error value includes a quadratic distance error value. Calculating the absolute value of the difference between two position errors at adjacent times to obtain the quadratic error value includes: calculating the absolute value of the difference between two distances at adjacent times to obtain the quadratic distance error value.
[0077] In some embodiments, the secondary error value includes a secondary angle error value. Calculating the absolute value of the difference between two position errors at adjacent times to obtain the secondary error value includes: calculating the absolute value of the difference between two angle differences at adjacent times to obtain the secondary angle error value.
[0078] In some embodiments, the quadratic error value includes a quadratic curvature error value. Calculating the absolute value of the difference between two position errors at adjacent times to obtain the quadratic error value includes: calculating the absolute value of the difference between two curvature differences at adjacent times to obtain the quadratic curvature error value.
[0079] When the second-order error value is greater than or equal to the preset difference threshold, the first lane centerline is processed based on the preset Kalman filter, which includes: when any one, two, or three of the second-order error values (second-order distance error value, second-order angle error value, and second-order curvature error value) are greater than or equal to the preset difference threshold, the first lane centerline is processed based on the preset Kalman filter.
[0080] In this embodiment, a Kalman filter is used to smooth the center line of the first lane, and then the position error is calculated based on the smoothed center line of the first lane and the center line of the second lane. This can obtain a more accurate and reliable position error, which is beneficial to improving the robustness of judging whether the lane has changed.
[0081] S44. Scoring is performed based on the position error corresponding to each sensing lane to obtain a confidence score for each sensing lane.
[0082] The confidence score represents the likelihood that a lane change will occur given a positional error. A higher confidence score indicates a higher probability of a lane change, while a lower confidence score indicates a lower probability of a lane change.
[0083] In one embodiment, step S44 includes: obtaining a first prior probability and a second prior probability, wherein the first prior probability is the prior probability when the actual lane corresponding to the first sensing lane changes, and the second prior probability is the prior probability when the actual lane corresponding to the first sensing lane does not change; inputting the position error corresponding to the first sensing lane into a preset first probability distribution model to obtain a first observation probability when the actual lane corresponding to the first sensing lane changes; inputting the position error corresponding to the first sensing lane into a preset second probability distribution model to obtain a second observation probability when the actual lane corresponding to the first sensing lane does not change; and inputting the first prior probability, the second prior probability, the first observation probability, and the second observation probability into a preset Bayesian model to obtain a confidence score for the first sensing lane.
[0084] Based on historical statistics or experience, a first prior probability P(change) is derived when the lane changes, and a second prior probability P(no change) is derived when the lane does not change. Error data of the lane centerline is collected for both lane change and no-change scenarios. Through analysis of a large amount of scenario data and fitting the error data using a probability distribution model (such as a Gaussian normal distribution), a first probability distribution model for "lane change" and a second probability distribution model for "lane no change" are obtained. The x-axis of both the first and second probability distribution models represents the error, and the y-axis represents the observation probability. When the position error is input as the x-axis to the first probability distribution model, the first probability distribution model outputs the observation probability corresponding to the position error as the x-axis, denoted as the first observation probability. When the position error is input as the x-axis to the second probability distribution model, the second probability distribution model outputs the observation probability corresponding to the position error as the x-axis, denoted as the second observation probability.
[0085] The confidence score expression based on the Bayesian model is: P(change|location error) = {f(location error|change)*P(change)} / {f(location error|change)*P(change) + f(location error|no change)*P(no change)}. Where P(change|location error) is the confidence score, f(location error|change) is the first observation probability, f(location error|no change) is the second observation probability, P(change) is the first prior probability, and P(no change) is the second prior probability.
[0086] For example, with a position error of 0.3, a first prior probability of 0.2, and a second prior probability of 0.8, inputting the position error of 0.3 into the first probability distribution model yields a first observation probability of 0.4. Inputting the position error of 0.3 into the second probability distribution model yields a second observation probability of 0.5. After the above calculations, the confidence score is approximately 0.167. Both the first and second prior probabilities are based on historical data of the error distribution or the experience of domain experts. By fusing these two prior probabilities and two observation probabilities through a Bayesian model, a more comprehensive assessment of the probability of position error occurring during lane changes can be achieved, which is more accurate and reliable than relying solely on the current observation probability.
[0087] S45 controls the vehicle to perform path planning based on confidence scores.
[0088] In one embodiment, step S45 includes adjusting the vehicle's navigation trajectory to obtain an updated navigation trajectory when the confidence score of the perceived lane corresponding to the actual lane in which the vehicle is located is greater than or equal to a preset score threshold.
[0089] In some embodiments, adjusting the vehicle's navigation trajectory to obtain an updated navigation trajectory includes: finding the lowest confidence score among multiple perceived lanes, and controlling the vehicle to change lanes to the real lane corresponding to the perceived lane with the lowest confidence score. When there are two or more perceived lanes with the lowest confidence score, the vehicle changes lanes to the real lane corresponding to one of the perceived lanes closest to the vehicle's current real lane.
[0090] In one embodiment, see Figure 8 Vehicle 100 needs to turn right from right-turn lane 82 of first road 81 to second road 83. However, the straight lane 84 in second road 83, which connects to right-turn lane 82, is blocked by a construction fence and has been diverted to straight lane 85. Vehicle 100 can detect straight lane 85 based on a perception map. When vehicle 100 turns right from right-turn lane 82 of first road 81 to straight lane 84 of second road 83, vehicle 100 detects that the confidence score of straight lane 84 is greater than a preset score threshold, therefore, straight lane 84 is changed. Next, vehicle 100 searches for the lowest confidence score among multiple perceived lanes, and the perceived lane with the lowest confidence score is straight lane 85. Vehicle 100 switches to straight lane 85 to drive, and the driving trajectory of vehicle 100 is shown by label 86.
[0091] Based on the vehicle control methods described in the above embodiments, the embodiments of this application have at least the following technical effects: They can predict in advance whether a lane change will occur, enabling the vehicle to plan a safe, reliable, and efficient path in a timely manner. Based on the positional error between the perceived center line of the first lane and the ideal center line of the second lane, the positional error of each perceived lane is scored to comprehensively assess the probability of a lane change. A lower confidence score indicates that the lane is less likely to change, while a higher confidence score indicates that the lane is more likely to change. Therefore, the confidence score clearly expresses the probability of a lane change, providing the vehicle with accurate and reliable decision-making information and helping the vehicle to anticipate and make more reliable and safer navigation strategies. By fusing multiple evaluation indicators to obtain the positional error, the positional error can accurately and reliably express the positional error between the center lines of the first and second lanes affected by obstacles of arbitrary shapes. This allows the system to adapt to various lane obstruction scenarios and maintain high reliability and accuracy in detecting lane changes.
[0092] It should be noted that in the above embodiments, there is no necessarily a certain order between the steps. Those skilled in the art can understand from the description of the embodiments of this application that the above steps may have different execution orders in different embodiments, that is, they may be executed in parallel or in turn, etc.
[0093] As another aspect of the embodiments of this application, this application provides a vehicle control device. The vehicle control device can be a software module, which includes several instructions stored in a memory. A processor can access the memory, invoke the instructions, and execute them to complete the vehicle control methods described in the various embodiments above.
[0094] In some embodiments, the vehicle control device can also be constructed from hardware components. For example, the vehicle control device can be constructed from one or more chips, which can work in coordination to complete the vehicle control methods described in the various embodiments above. As another example, the vehicle control device can also be constructed from various logic devices, such as general-purpose processors, digital signal processors, application-specific integrated circuits, field-programmable gate arrays, microcontrollers or other programmable logic devices, or any combination of these components.
[0095] In one embodiment, see Figure 9 The vehicle control device 900 includes a first lane determination module 91, a second lane determination module 92, a position error determination module 93, a confidence score determination module 94, and a vehicle control module 95. The first lane determination module 91 determines the spatial position of the center line of the first lane corresponding to each perceived lane based on perception data collected by sensors on the vehicle, wherein one perceived lane corresponds to a real lane on the driving road. The second lane determination module 92 acquires a high-precision map and determines the spatial position of the center line of the second lane corresponding to each perceived lane based on the high-precision map. The position error determination module 93 determines the position error corresponding to each perceived lane based on the spatial positions of the first and second lane center lines. The confidence score determination module 94 scores each perceived lane based on its position error to obtain a confidence score for each perceived lane. The vehicle control module 95 controls the vehicle to perform path planning based on the confidence score.
[0096] In some embodiments, the sensors on the vehicle include LiDAR and cameras. The first lane line determination module 91 is specifically used to: perform data fusion based on the point cloud data collected by the LiDAR and the image data collected by the camera to obtain a perception map corresponding to the driving road, input the perception map into a pre-trained road structure perception model, and obtain the spatial position of the first lane centerline corresponding to each perception lane.
[0097] In some embodiments, the high-precision map includes the spatial location of at least one standard lane, and one standard lane corresponds to one real lane. The second lane line determination module 92 is specifically used to: obtain the current pose of the vehicle, search for the standard lane corresponding to each sensing lane in the high-precision map according to the current pose of the vehicle and preset filtering conditions, and use the spatial location of the lane center line of the standard lane corresponding to each sensing lane as the spatial location of the second lane center line corresponding to each sensing lane.
[0098] In some embodiments, the driving road includes at least one real lane, the at least one real lane includes a first real lane, and the sensing lane corresponding to the first real lane includes a first sensing lane; the position error determination module 93 is specifically used to: map the spatial position of the center line of the first lane corresponding to the first sensing lane to the world coordinate system to obtain the spatial position of the center line of the third lane corresponding to the first sensing lane; map the spatial position of the center line of the second lane corresponding to the first sensing lane to the world coordinate system to obtain the spatial position of the center line of the fourth lane corresponding to the first sensing lane; obtain at least one error evaluation index based on the spatial position of the center line of the third lane corresponding to the first sensing lane and the spatial position of the center line of the fourth lane corresponding to the first sensing lane; and obtain the position error corresponding to the first sensing lane based on the at least one error evaluation index.
[0099] In some embodiments, at least one error evaluation index includes a first type of evaluation index. The position error determination module 93 is specifically used to: perform discrete point sampling on the center line of the third lane corresponding to the first sensing lane to obtain multiple current position points; perform discrete point sampling on the center line of the fourth lane corresponding to the first sensing lane to obtain multiple standard position points, wherein one standard position point corresponds to one current position point; calculate a first distance value between each current position point and each standard position point corresponding to the current position point; and generate a first type of evaluation index based on the multiple first distance values, wherein the first type of evaluation index is used to evaluate the distance error between the center line of the third lane corresponding to the first sensing lane and the center line of the fourth lane corresponding to the first sensing lane.
[0100] In some embodiments, the error evaluation index further includes a second type of evaluation index. The position error determination module 93 is specifically used to: calculate multiple angle differences, wherein the angle difference is the absolute value of the difference between the first tangent angle of the current position point and the second tangent angle of the standard position point corresponding to the current position point; generate a second type of evaluation index based on the multiple angle differences, the second type of evaluation index being used to evaluate the angle error between the center line of the third lane corresponding to the first sensing lane and the center line of the fourth lane corresponding to the first sensing lane.
[0101] In some embodiments, the error evaluation index further includes a third type of evaluation index. The position error determination module 93 is specifically used to: calculate multiple curvature difference values, wherein the curvature difference value is the absolute value of the difference between the first curvature of the current position point and the second curvature of the standard position point corresponding to the current position point; generate a third type of evaluation index based on the multiple curvature difference values, the third type of evaluation index being used to evaluate the curvature error between the center line of the third lane corresponding to the first sensing lane and the center line of the fourth lane corresponding to the first sensing lane.
[0102] In some embodiments, the driving road includes at least one real lane, the at least one real lane includes a first real lane, and the sensing lane corresponding to the first real lane includes the first sensing lane; the confidence determination module 94 is specifically used to: obtain a first prior probability and a second prior probability, the first prior probability being the prior probability when the real lane corresponding to the first sensing lane changes, and the second prior probability being the prior probability when the real lane corresponding to the first sensing lane does not change; input the position error corresponding to the first sensing lane into a preset first probability distribution model to obtain a first observation probability when the real lane corresponding to the first sensing lane changes; input the position error corresponding to the first sensing lane into a preset second probability distribution model to obtain a second observation probability when the real lane corresponding to the first sensing lane does not change; input the first prior probability, the second prior probability, the first observation probability, and the second observation probability into a preset Bayesian model to obtain a confidence score for the first sensing lane.
[0103] In some embodiments, the vehicle control module 95 is specifically used to: adjust the vehicle's navigation trajectory to obtain an updated navigation trajectory when the confidence score of the perceived lane corresponding to the actual lane where the vehicle is located is greater than or equal to a preset score threshold.
[0104] See Figure 10 , Figure 10 This is a schematic diagram of a controller provided in an embodiment of this application. The controller 101 includes one or more processors 102 and a memory 103. The memory 103 is connected to one or more processors 102, for example, via a bus.
[0105] Processor 102 is configured to support the controller in performing the corresponding functions in the methods described in the above method embodiments. The processor may be a Central Processing Unit (CPU), a Network Processor (NP), a hardware chip, or any combination thereof. The hardware chip may be an Application Specific Integrated Circuit (ASIC), a Programmable Logic Device (PLD), or a combination thereof. The PLD may be a Complex Programmable Logic Device (CPLD), a Field-Programmable Gate Array (FPGA), a Generic Array Logic (GAL), or any combination thereof.
[0106] Memory 103 is used to store program code, etc. Memory may include volatile memory (VM), such as random access memory (RAM); memory may also include non-volatile memory (NVM), such as read-only memory (ROM), flash memory, hard disk drive (HDD), or solid-state drive (SSD); memory may also include combinations of the above types of memory.
[0107] The memory 103 can be used to store non-volatile software programs, non-volatile computer-executable programs, and modules, such as the program instructions / modules corresponding to the vehicle control method in the embodiments of this application. The processor executes various functional applications and data processing of the vehicle control method and vehicle control device by running the non-volatile software programs, instructions, and modules stored in the memory, that is, it realizes the functions of the various modules or units of the vehicle control method and vehicle control device provided in the above method embodiments.
[0108] The memory may include a program storage area and a data storage area, wherein the program storage area may store the operating system and application programs required for at least one function. The data storage area may store data created based on the use of the vehicle control device. In some embodiments, the memory may optionally include memory remotely located relative to the processor, which can be connected to the vehicle control device via a network. Examples of such networks include, but are not limited to, the Internet, intranets, local area networks, mobile communication networks, and combinations thereof.
[0109] The one or more modules are stored in the memory. When executed by the one or more processors, they perform the vehicle control method in any of the above method embodiments. For example, they perform the method steps described in the above method embodiments to realize the functions of the modules described in the above device embodiments.
[0110] This application also provides a computer-readable storage medium storing a computer program, the computer program including program instructions, which, when executed by a computer, cause the computer to perform the method described in the foregoing embodiments.
[0111] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. The storage medium can be a magnetic disk, optical disk, read-only memory (ROM), or random access memory (RAM), etc.
[0112] The above-disclosed embodiments are merely preferred embodiments of this application and should not be construed as limiting the scope of this application. Therefore, any equivalent variations made in accordance with the claims of this application shall still fall within the scope of this application.
Claims
1. A vehicle control method, characterized in that, include: Based on the perception data collected by the sensors on the vehicle, the spatial position of the first lane centerline corresponding to each perception lane is obtained, wherein one perception lane corresponds to one actual lane on the driving road. Acquire a high-precision map, and based on the high-precision map, obtain the spatial position of the center line of the second lane corresponding to each of the sensing lanes; Based on the spatial position of the first lane centerline corresponding to each of the sensing lanes and the spatial position of the second lane centerline corresponding to each of the sensing lanes, the position error corresponding to each of the sensing lanes is obtained. A confidence score for each sensing lane is obtained by scoring the position error corresponding to each sensing lane. The vehicle is controlled to perform path planning based on the confidence score.
2. The vehicle control method according to claim 1, wherein the sensors on the vehicle include lidar and cameras, characterized in that, The step of obtaining the spatial position of the first lane centerline corresponding to each sensing lane based on the sensing data collected by the sensors on the vehicle includes: The point cloud data collected by the lidar and the image data collected by the camera are fused to obtain a perception map corresponding to the driving road. The perception map is input into a pre-trained road structure perception model to obtain the spatial position of the first lane centerline corresponding to each perception lane.
3. The vehicle control method according to claim 1, characterized in that, The high-precision map includes the spatial location of at least one standard lane, and one standard lane corresponds to one actual lane. The step of obtaining the spatial position of the second lane centerline corresponding to each of the sensing lanes based on the high-precision map includes: Obtain the current pose of the vehicle; Based on the vehicle's current position and preset filtering conditions, the standard lane corresponding to each of the sensing lanes is searched in the high-precision map. The spatial position of the center line of the standard lane corresponding to each sensing lane is taken as the spatial position of the second lane center line corresponding to each sensing lane.
4. The vehicle control method according to claim 1, characterized in that, The driving road includes at least one of the real lanes, the at least one of the real lanes includes a first real lane, and the sensing lane corresponding to the first real lane includes a first sensing lane; The step of obtaining the position error corresponding to each sensing lane based on the spatial position of the first lane centerline corresponding to each sensing lane and the spatial position of the second lane centerline corresponding to each sensing lane includes: Map the spatial position of the center line of the first lane corresponding to the first sensing lane to the world coordinate system to obtain the spatial position of the center line of the third lane corresponding to the first sensing lane. Map the spatial position of the center line of the second lane corresponding to the first sensing lane to the world coordinate system to obtain the spatial position of the center line of the fourth lane corresponding to the first sensing lane. Based on the spatial position of the center line of the third lane corresponding to the first sensing lane and the spatial position of the center line of the fourth lane corresponding to the first sensing lane, at least one error evaluation index is obtained. The position error corresponding to the first sensing lane is obtained based on at least one of the error evaluation indicators.
5. The vehicle control method according to claim 4, characterized in that, At least one of the error evaluation indicators includes a first type of evaluation indicator. The step of obtaining at least one error evaluation indicator based on the spatial position of the center line of the third lane corresponding to the first sensing lane and the spatial position of the center line of the fourth lane corresponding to the first sensing lane includes: Discrete point sampling is performed on the centerline of the third lane corresponding to the first sensing lane to obtain multiple current position points; Discrete point sampling is performed on the center line of the fourth lane corresponding to the first sensing lane to obtain multiple standard position points, wherein one of the standard position points corresponds to one of the current position points; Calculate the first distance value between each current location point and the corresponding standard location point; The first type of evaluation index is generated based on multiple first distance values, wherein the first type of evaluation index is used to evaluate the distance error between the center line of the third lane corresponding to the first sensing lane and the center line of the fourth lane corresponding to the first sensing lane.
6. The vehicle control method according to claim 5, characterized in that, The error evaluation index further includes a second type of evaluation index, and the method further includes: Calculate multiple angle differences, wherein the angle difference is the absolute value of the difference between the first tangent angle of the current position point and the second tangent angle of the standard position point corresponding to the current position point; The second type of evaluation index is generated based on multiple angle differences. The second type of evaluation index is used to evaluate the angle error between the center line of the third lane corresponding to the first sensing lane and the center line of the fourth lane corresponding to the first sensing lane.
7. The vehicle control method according to claim 5, characterized in that, The error evaluation index also includes a third type of evaluation index, and the method further includes: Calculate multiple curvature differences, wherein the curvature difference is the absolute value of the difference between the first curvature of the current position point and the second curvature of the standard position point corresponding to the current position point; The third type of evaluation index is generated based on multiple curvature differences. The third type of evaluation index is used to evaluate the curvature error between the center line of the third lane corresponding to the first sensing lane and the center line of the fourth lane corresponding to the first sensing lane.
8. The vehicle control method according to claim 1, characterized in that, The driving road includes at least one of the real lanes, and the at least one of the real lanes includes a first real lane. The sensing lane corresponding to the first real lane includes the first sensing lane. The step of scoring each sensing lane based on its corresponding position error to obtain a confidence score for each sensing lane includes: Obtain a first prior probability and a second prior probability. The first prior probability is the prior probability when the actual lane corresponding to the first perceived lane changes, and the second prior probability is the prior probability when the actual lane corresponding to the first perceived lane does not change. Input the position error corresponding to the first perceived lane into a preset first probability distribution model to obtain the first observation probability when the actual lane corresponding to the first perceived lane changes; Input the position error corresponding to the first perceived lane into a preset second probability distribution model to obtain the second observation probability when the actual lane corresponding to the first perceived lane has not changed. The first prior probability, the second prior probability, the first observation probability, and the second observation probability are input into a preset Bayesian model to obtain the confidence score of the first sensing lane.
9. The vehicle control method according to claim 1, characterized in that, The step of controlling the vehicle to perform path planning based on the confidence score includes: When the confidence score of the perceived lane corresponding to the actual lane where the vehicle is located is greater than or equal to a preset score threshold, the navigation trajectory of the vehicle is adjusted to obtain an updated navigation trajectory.
10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program, the computer program including program instructions that, when executed by a processor, cause the processor to perform the vehicle control method as described in any one of claims 1-9.
