JP2023088593A - Method and system for extracting road surface data, and method and system for controlling automatic driving vehicle - Google Patents

Abstract

To provide a method and system for acquiring road surface data, and a method and system for controlling an automatic driving vehicle.SOLUTION: A method for extracting road surface data includes: acquiring the road surface data from first environment detection data of a detection area; acquiring a first road surface model using the road surface data; acquiring a second road surface model by complementing a deficient area of the first road surface model; acquiring road surface attributes of the detection area according to the second road surface model; identifying road surface damage using second environment detection data of the detection area and the second road surface model; and storing the road surface attributes and a result of the road surface damage identification as road surface data of the detection area.SELECTED DRAWING: Figure 2

Description

TECHNICAL FIELD The present disclosure relates to methods for extracting road surface data, and more particularly to methods for extracting road surface data, including building a ground model.

With advances in artificial intelligence and related technologies, self-driving cars are being recognized as the future of transportation. The global automotive industry is investing in developing the field of autonomous driving. A self-driving car can program an appropriate driving route according to the current situation of the self-driving car and road conditions. Therefore, acquiring sophisticated road surface data is one of the important issues in the field of autonomous driving.

With the rapid development of urban areas and the improvement of computer software technology, there is a growing demand for acquisition and analysis of road surface data not only in the field of autonomous driving, but also in areas such as planning and maintenance of urban redevelopment, car navigation, and location information services. increasing.

In view of the above description, the present disclosure provides methods and systems for acquiring road surface data and methods and systems for controlling self-driving vehicles.

According to one embodiment of the present disclosure, a method for obtaining road surface data includes executing the following steps in a processing device: obtaining ground data from first environment detection data in a detection area; obtaining a first ground model from the ground data; filling missing areas of the first ground model to obtain a second ground model; obtaining a plurality of road surface attributes of the detection area according to the second ground model; obtaining; performing road surface damage identification using the second environment detection data of the detection area and the second ground model; and storing the plurality of road surface attributes and road surface damage identification results as a plurality of road surface data of the detection area. to do.

According to one embodiment of the present disclosure, a method for controlling an autonomous vehicle includes executing the following steps in a processing device: according to the current position and speed of the vehicle, after a processing time and a vehicle chassis control delay time; obtaining target road surface data corresponding to the estimated position from a map database; and generating steering commands to control the vehicle according to the target road surface data. The map database comprises a plurality of pre-stored road surface data of the detection area, the estimated position is located in the detection area, the target road surface data is within the plurality of pre-stored road surface data, the pre-stored road surface data A plurality of road surface data are acquired by road surface data extraction processing. The road surface data extraction processing includes obtaining ground data from the first environment detection data in the detection area, obtaining a first ground model from the ground data, compensating for the missing area of the first ground model and obtaining a second ground model. obtaining a plurality of road surface attributes of the detection area according to the second ground model; and performing road surface damage identification using the second environment detection data of the detection area and the second ground model. and storing the plurality of road surface attributes and road damage identification results as a plurality of pre-stored road surface data in the detection area.

According to one embodiment of the present disclosure, a system for road surface data extraction includes a data input device, a storage device, and a processing device, the processing device being connected to the data input device and the storage device. The data input device is configured to receive first environmental sensing data and second environmental sensing data. The processing device is configured to perform the following steps: obtaining ground data from the first environment detection data; obtaining a first ground model using the ground data; obtaining the first ground model. obtaining a second ground model by compensating for the missing area of the second ground model; obtaining a plurality of road surface attributes of the detection area based on the second ground model; combining the second environment detection data with the second ground model performing road surface damage identification using a plurality of road surface attributes and road surface damage identification results as a plurality of road surface data in the detection area.

According to one embodiment of the present disclosure, a system for controlling an autonomous vehicle comprises a positioning module, a vehicle dynamic state sensing module, a steering angle control module, and a processing module, the processing module comprising: a positioning module; It is connected to the vehicle dynamic state sensing module and the steering angle control module. The positioning module is configured to obtain the current position of the vehicle. The vehicle dynamic state sensing module is configured to obtain the current speed of the vehicle. The steering angle control module is configured to control the vehicle according to the steering commands. The processing module is configured to perform the following steps: estimating an estimated position after the processing time and the vehicle chassis control delay time according to the current position and speed of the vehicle; determining a target road surface corresponding to the position; retrieving data from a map database; generating steering commands to control the vehicle according to the target road surface data. The map database comprises a plurality of pre-stored road surface data of the detection area, the estimated position is located in the detection area, the target road surface data is within the plurality of pre-stored road surface data, the pre-stored road surface data A plurality of road surface data are acquired by road surface data extraction processing. The road surface data extraction processing includes acquiring ground data from the first environment detection data in the detection area, acquiring a first ground model using the ground data, and filling the missing area of the first ground model. obtaining a second ground model; obtaining a plurality of road surface attributes of the detection area according to the second ground model; and using the second environment detection data of the detection area and the second ground model to identify road damage. Storing the plurality of road surface attributes and road surface damage identification results as a plurality of pre-stored road surface data in the detection area.

The above content and the following detailed description of the disclosure are used to illustrate the concept and spirit of the disclosure and to provide further explanation of the scope of the disclosure.

1 is a functional block diagram of an operating environment of a system for extracting road surface data according to one embodiment of the present disclosure; FIG. 4 is a flowchart of a method for extracting road surface data, according to one embodiment of the present disclosure; 4 is a flowchart of steps for obtaining ground surface data in a method for extracting road surface data, according to one embodiment of the present disclosure; FIG. 4 is a schematic diagram of performing the step of obtaining ground surface data in a method for extracting road surface data, according to one embodiment of the present disclosure; 4 is a flow chart of the steps of filling missing areas in a method for extracting road surface data, according to one embodiment of the present disclosure. FIG. 3 is a schematic diagram of environment detection data in a detection area, a first ground model, and a second ground model, according to one embodiment of the present disclosure; 4 is a flowchart of road damage identification steps in a method for extracting road surface data, according to one embodiment of the present disclosure. FIG. 4 is a schematic diagram of performing the step of road damage identification in a method for extracting road surface data, according to one embodiment of the present disclosure; FIG. 4 is a schematic diagram of performing the step of road damage identification in a method for extracting road surface data, according to one embodiment of the present disclosure; FIG. 4 is a schematic diagram of performing the step of road damage identification in a method for extracting road surface data, according to one embodiment of the present disclosure; 1 is a functional block diagram of an operating environment of a system for controlling an autonomous vehicle, according to one embodiment of the present disclosure; FIG. 1 is a schematic diagram of a method for controlling a self-driving vehicle performed by a system for controlling a self-driving vehicle, according to one embodiment of the present disclosure; FIG. 4 is a flowchart of a method for controlling an autonomous vehicle, according to one embodiment of the present disclosure;

In the following detailed description, for purposes of explanation, numerous specific details are set forth in order to provide a thorough understanding of the disclosed embodiments. The concepts and features of the present invention can be readily understood by those skilled in the art from the description, claims, and drawings disclosed herein. The following embodiments further illustrate various aspects of the invention, but are not intended to limit the scope of the invention.

Please refer to FIG. 1, which is a functional block diagram of the operating environment of a system for extracting road surface data, according to one embodiment of the present disclosure. The system 1 for extracting road surface data can obtain the environment detection data of a certain detection area, and consequently build a ground model of the detection area based on the detection data of the detection area. A system 1 for extracting road surface data performs complementary operations on the ground model to optimize the ground model, and extracts multiple road surface attributes (e.g., slope angle, slope, radius of curvature, etc.) and road surface conditions (e.g., , potholes, ruts, etc.) can be obtained. As shown in FIG. 1, a system 1 for extracting road surface data comprises a data input device 11, a storage device 13 and a processing device 15, the processing device 15 being connected to the data input device 11 through a wired or wireless connection. It is connected to device 13 .

The data input device 11 may comprise a wireless communication module using communication technologies such as Wi-Fi, 4G, 5G. The data input device 11 is connected to the environment detector 2 and configured to receive environment sensing data detected by the environment detector 2 . Although FIG. 1 shows a typical example with one environment detector 2 , the data input device 11 may be connected to multiple environment detectors 2 . The environment detector 2 may be a LIDAR, and environment detection data to be detected may be a point cloud. The environment detector 2 may be located at a fixed location or mounted on a moving platform such as an aircraft, unmanned aerial vehicle (UAV) or vehicle. Here, the vehicle is, for example, a radar plot vehicle, an automatic driving vehicle, or the like. The data input device 11 provides the environment detection data acquired by one or more environment detectors 2 to the processing device 15 for extraction of road surface data of a specific detection area. The particular detection area depends on the detection range of the one or more environmental detectors 2 .

Storage device 13 may comprise, but is not limited to, flash memory, hard disk drive (HDD), solid state disk (SSD), dynamic random access memory (DRAM), static random access memory (SRAM). The storage device 13 can store road surface data extracted by the processing device 15 as described above. In one embodiment, storage device 13, data input device 11, and processing device 15 are components of the same server. In other embodiments, storage device 13 and processing device 15 are provided in different servers and are connected to each other by wired or wireless connections.

Processing unit 15 is a single processor or collection of multiple processors, such as, but not limited to, a central processing unit (CPU) or graphics processing unit (GPU). The processing device 15 acquires environment detection data of the detection area from the data input device 11, processes the environment detection data to extract road surface data of the detection area, and stores or stores the extracted road surface data in the storage device 13. It is arranged to update the road surface data stored in the device 13 . The road surface data may include road surface attributes and the results of road surface damage identification, and implementing methods for extracting the road surface data are described below.

See FIGS. 1 and 2. FIG. 2, which is a flowchart of a method for extracting road surface data according to one embodiment of the present disclosure. As shown in FIG. 2, a method for extracting road surface data includes step S11: obtaining ground surface data from the first environment detection data in the detection area; Obtaining a model, Step S13: Filling in missing areas of the first ground model to obtain a second ground model, Step S14: Obtaining a plurality of attributes of the detection area according to the second ground model. Step S15: performing road surface damage identification using the second environment detection data of the detection region and the second ground model; Including storing as data. It should be mentioned that the present disclosure does not limit the order of steps S14 and S15. The method for extracting road surface data shown in FIG. 2 can, but is not limited to, be performed by the processing device 15 as shown by the system 1 for extracting road surface data. For ease of understanding, the steps of the method for extracting road surface data will be described using the operation by the processing device 15 as an example.

In step S11, the processing device 15 can acquire the first environment detection data of the detection area from the environment detector 2 via the data input device 11, and acquire the ground data from the first environment detection data. More specifically, see FIGS. 3 and 4. FIG. 3, wherein FIG. 3 is a flowchart of steps of obtaining ground surface data in a method for extracting road surface data according to one embodiment of the present disclosure, and FIG. 4 is a flow chart of one embodiment of the present disclosure. 1 is a schematic diagram of performing a step of obtaining ground surface data in a method for extracting road surface data according to FIG. As shown in FIG. 3, the step of obtaining ground data includes sub-step S111: dividing the first environment detection data (point cloud) into a plurality of original voxels, where the plurality of original voxels is a plurality of voxels forming columns, each of the plurality of voxel columns extending along a vertical axis, sub-step S112: obtaining a plurality of target voxels from the plurality of original voxels, wherein the plurality of target voxels are respectively In each of the plurality of target voxels, the surface normal of the plurality of points in the plurality of voxel columns is upward, and the vertical diffusion parameter of the plurality of points is less than or equal to a predetermined value, substep S113: the plurality of targets calculating a plurality of elevation values for the voxels, where each of the plurality of elevation values represents the average value of the vertical axis values of the plurality of points within the plurality of target voxels; and substep S114. : includes generating ground data using multiple altitude values.

Each of the substeps is further described. In substeps S111 and S113, the vertical axis may be the axis perpendicular to the horizontal plane. A sub-step S112 obtains a plurality of vertically continuous voxel groups from each voxel column in each of the plurality of voxel columns, and recognizes the lowest voxel group satisfying a plurality of specific conditions as a target voxel group. may include The specific conditions are (1) that the normal to the surface formed by a plurality of points included in the continuous voxel group is upward, and (2) that the vertical diffusion parameters of the points included in the continuous voxel group are predetermined. may include being less than or equal to the value of The lowest voxel group satisfying the specified condition may be the one closest to the horizontal plane on the vertical axis among one or more consecutive voxel groups satisfying conditions (1) and (2) above. The surface normal (surface normal) in condition (1) may be determined by principal component analysis and scanning trajectory. For example, the scan trajectory may be the scan trajectory of the environment detector that generated the environment detection data. Specifically, the principal component analysis may estimate two normal candidates in opposite directions, where the normal candidate lying on the same side of the scan trajectory is selected as the surface normal. The fact that the surface normal is upwards means that the angle between the surface normal and the vertical axis is less than the threshold. The vertical diffusion parameter in condition (2) represents the distance between points distributed along the vertical axis mentioned above. A predetermined value is, for example, 30 cm. In one embodiment, processor 15 may delete points outside the target voxels, ie retain points within the target voxels.

Sub-step S114 performs a region growing operation or a connecting quantification operation on a plurality of altitude values obtained from a plurality of target voxel groups to form a plurality of altitude value connecting groups, and a scanning trajectory and one having overlap. Selecting one or more altitude value connections as ground data may be included.

Take the point cloud PC shown in FIG. 4 as an example. A voxel column VC is composed of the original voxels OV arranged along the z-axis. Although FIG. 4 shows only one voxel train VC composed of the original voxels OV and the target voxel group TV included in the voxel train VC, in reality, all the point clouds PC are divided into a plurality of original voxels OV. , and the original voxels OV may constitute a plurality of voxel columns VC. The number of voxel columns VC and the number of original voxels OV that make up each voxel column are not limited to the present disclosure. In FIG. 4, the voxel column VC is composed of eight original voxels OV, labeled voxels 1-8 from bottom to top. Voxels 1 and 4-7 contain points. Voxels 2, 3 and 8 do not contain points. Voxels 1 and 4 to 7 can be divided into two contiguous voxel groups constituted by voxel column 1 and voxel column 4 to 7 respectively. A point in the contiguous group of voxels constituted by voxel column 1 satisfies the condition that the surface normal is upward and the vertical diffusion parameter is less than a predetermined value, thus the contiguous group of voxels constituted by voxel column 1 The voxel group serves as the target voxel group TV. The mean value of the values on the vertical axis z of the points in the target voxel TV serves as the altitude value for the target voxel TV and is used to generate the ground data. The method of generating the ground data is as described above and will not be repeated here.

Please refer to FIGS. 1 and 2 again. At step S12, the processing device 15 uses the ground data to obtain a first ground model. In one embodiment, the processor 15 uses the ground data acquired in step S11 as the first ground model. In another embodiment, processing unit 15 obtains a first ground model by approximating the ground data using a curved surface model. In step S13, the processing device 15 fills in the missing area in the first ground model to acquire the second ground model. More specifically, please refer to FIG. 5, which is a flowchart of the steps of filling missing areas in a method for extracting road surface data, according to one embodiment of the present disclosure. As shown in FIG. 5, the step for filling the missing area includes sub-step S131: storing the first ground model in two storage units each comprising a plurality of storage units respectively storing a plurality of altitude values of the first ground model. converting to dimensional data; sub-step S132: performing digital inpainting on the two-dimensional data to generate inpainted two-dimensional data; and sub-step S133: applying the inpainted two-dimensional data to a second ground model. can include converting to a three-dimensional model as The two-dimensional data may be regular grids, for example. A missing region may be a blocked region or a non-detected region. Digital restoration can fill in missing regions with arbitrary shapes. For example, the digital inpainting may solve the function f of the defect region Ω, and the digital inpainting may minimize the Laplacian function Δf while maintaining the continuity of the defect region boundary ∂Ω. The function f and the Laplacian function Δf are shown as follows.

Please refer to FIG. 6, which is a schematic diagram of the environment detection data of the detection area, the first ground model, and the second ground model, according to one embodiment of the present disclosure. FIG. 6 shows the environment detection data (point cloud) of the detection areas A to C. Here, the first ground model uses the above step S11 and step S12 in FIG. , and the second ground model is obtained by performing step S13 in FIG. 2 on the first ground model corresponding to the detection area AC.

Please refer to FIGS. 1 and 2 again. In step S14, the processing device 15 acquires road surface attributes of the detection area according to the second ground model. More specifically, the processing unit 15 obtains from the second ground model a plurality of altitude values within an area formed by a predetermined driving route that extends a predetermined distance outward, and the (road surface) of the predetermined driving route. It can calculate road surface attributes such as slope, slope and radius of curvature of the surrounding terrain. There may be one or more predetermined driving routes. For example, the predetermined driving route may be a Google Maps suggested route or a pre-programmed driving route by the user, and the predetermined distance may be a typical paved road or the width of the vehicle, but the predetermined driving route and the predetermined distance are not limited to these.

In step S15, the processing device 15 performs road damage identification using the second environment detection data of the detection area and the second road surface model. Points in the initially detected point cloud that belong to road damage (e.g. potholes and ruts) may be determined as non-ground data in step S11 and not included in the ground model. It is first done using the detected point cloud and the ground model, filling in the data of the road damage locations. In one embodiment, the second environment detection data and the first environment detection data for generating the second road surface model belong to the same detection area but are detected at different times. Since the rough surface damage position may change after a certain period of time, the processing device 15 constructs the road surface model and after a certain period of time has passed, the data input device 11 transmits the second environment detection data to the environment detector 2 . to perform road damage identification, and then update the ground data corresponding to the ground model. In particular, the processing device may periodically acquire environment detection data detected by the environment detector 2 and periodically update the road surface data.

Step S15 will be further described. Please refer to FIG. 7, which is a flowchart of road damage identification steps in a method for extracting road surface data, according to one embodiment of the present disclosure. As described above, the first road surface model may be the road surface data obtained in step S11, or may be a curved surface model approximated to the road surface data. obtained by doing Therefore, the second road surface model may have or be a curved surface (hereinafter referred to as the original curved surface). As shown in FIG. 7, the steps of road damage identification include substep S151: shifting the original curved surface down a first distance along the vertical axis to create a first curved surface; down a second distance that is greater than the first distance along the vertical axis to create a second surface; and substep S153: shifting the first surface and the first It may include determining a plurality of points between the two curved surfaces to form one or more damage regions. In particular, the first surface can be related to the accuracy of the point cloud, which depends on the accuracy of the scan of the environment detector; ) may be related to detection errors produced by environmental detectors due to certain environmental factors. For example, the second distance may be set to 6 cm if the value on the vertical axis of the water reflectance data on the road surface is typically less than 6 cm. Note that the order of sub-step S151 and sub-step S152 is not limited in the present disclosure.

The sub-steps described above will be explained with schematic diagrams. Please refer to FIGS. 8-10, which are schematic diagrams of performing the steps of road damage identification in a method for extracting road surface data, according to one embodiment of the present disclosure. FIG. 8(a) exemplarily shows the second environment detection data, and an enlarged view of the area B1 is shown in FIG. 8(b). FIG. 9 exemplarily shows end views of the second environment detection data, the original curved surface C0, the first curved surface C1, and the second curved surface C2 cut by the cutting line SL1. A first surface C1 is obtained by shifting the original surface C1 along the vertical axis z down a first distance D1, and a second surface C2 is obtained by shifting the original surface C0 along the vertical axis z by a second distance D1. 2 by shifting down a distance D2. In this embodiment, the first distance D1 is 3 cm and the second distance D1 is 6 cm, but the disclosure is not so limited. As shown in FIG. 10(b), points between the first curve C1 and the second curve C2 in the second environment detection data (point group) are determined as points forming the damaged region DR1. . FIGS. 10(a) and 10(b) show data schematic diagrams of marked road damage locations in the second environment detection data shown in FIGS. 8(a) and 8(b), respectively.

Please refer to FIGS. 1 and 2 again. In step S16, the processing device 15 stores the road surface attribute and the road surface damage identification result as the road surface data of the detection area in the storage device. Specifically, road surface attributes and road damage identification results can be stored in one of the following formats: irregular triangulation, regular grid, and waypoints. The waypoints may be a plurality of points on the predetermined driving path, each of the points on the predetermined driving path having road surface attributes and/or road damage identification results (e.g., damaged/undamaged) at the corresponding location. ). In one embodiment, the road surface data may further include a second road surface model, which may be stored in one of the above formats along with the road surface attributes and road damage identification results. In particular, data such as road surface model, road surface attributes, and road damage identification results can be stored in one of the following formats: irregular triangulation, regular grid, and waypoints. One or more of the above data may be stored as the same or different data, and the disclosure is not so limited.

In particular, the method for extracting road surface data and the system for extracting road surface data described in the above one or more embodiments apply to building high-definition maps and/or controlling self-driving vehicles. be able to.

Please refer to FIG. 11, which is a functional block diagram of the operating environment of a system for controlling an autonomous vehicle, according to one embodiment of the present disclosure. As shown in FIG. 11 , the system for controlling the autonomous vehicle 3 comprises a positioning module 31 , a vehicle dynamic state detection module 33 , a processing module 35 and a steering angle control module 37 . The processing module 35 may be wired or wirelessly connected to the positioning module 31 , the vehicle dynamic state sensing module 33 and the steering angle module 37 , and the positioning module 31 and the processing module 35 are wired or wirelessly connected to the map database 4 . may be connected. The map database 4 includes maps of one or more detection areas each containing pre-stored road surface data. The details are not described again, as they can be obtained by the method for and the system for extracting the road surface data. In particular, the map database 4 can be realized by the storage device in the system for extracting road surface data described in the above embodiments.

The positioning module 31 is configured to obtain the current position of the vehicle to be controlled and may include, but is not limited to, position sensors or the like. The vehicle dynamic state sensing module 33 is configured to obtain the current speed of the vehicle and may comprise dynamic sensors such as wheel speed sensor, steering wheel angle sensor, yaw angle sensor, deceleration sensor, etc., which include: Not limited. The steering angle control module 37 is configured to control the steering angle of the vehicle according to the steering command, and controls the braking force and driving force of the vehicle, such as the throttle body, solenoid valve, and relay, to control the yaw motion of the vehicle. elements can be provided, but are not limited to these. The processing module 35 can be, for example, a microcontroller in an electronic control unit (ECU) or a central processing unit in a microcontroller, and according to the current position and current speed of the vehicle being controlled, the processing time and the vehicle chassis control delay time. It is configured to estimate a later position, acquire target road surface data corresponding to the estimated position from the map database 4, generate a steering command according to the target road surface data, and control the vehicle via the steering angle control module 37. be done. The estimated position is included in one of the one or more detection regions. The processing module 35 can search the map database 4 for road surface data of the detection area where the estimated position is located, and obtain target road surface data for the estimated position from the road surface data.

The operation of the system for controlling the autonomous vehicle 3 is further described. Please refer to FIG. 12, which is a schematic illustration of performing a method for controlling a self-driving vehicle by a system for controlling a self-driving vehicle, according to one embodiment of the present disclosure. The processing module 35 may comprise a travel distance estimation sub-module 351 , a road surface data collection sub-module 353 and a route tracking sub-module 355 . For example, the above sub-modules may be implemented by one or more instructions or programs each having a function of estimating a distance traveled, a function of collecting road surface data, and a function of tracking a route.

In communication operations A 11 and A 12 , the positioning module 31 obtains the coordinates of the current position of the vehicle on the map and provides the coordinates of the current position to the road surface data collection sub-module 353 . In communication operations A13 and A14, the vehicle dynamic state detection module 33 detects vehicle dynamic information (such as current speed, current yaw angular velocity, etc.) from the vehicle body 5, and transmits the vehicle dynamic information to the movement distance estimation sub-module. provided to module 351; The movement distance estimation sub-module 351 obtains the total delay time by adding the processing time of the processing module 35 and the vehicle chassis control delay time of the vehicle body 5, and multiplies the total delay time by the current speed to estimate the movement. Obtain the distance and transmit the estimated traveled distance to the route tracking sub-module 355 in communication operation A15. For example, the processing time of the processing module 35 is the total estimated time of performing data processing by the travel distance estimation sub-module 351 , the road surface data collection sub-module 353 and the route tracking sub-module 355 . The road surface data collection sub-module 353 obtains the coordinates of the estimated position using the current position and the estimated moving distance. In communication operations A16-A18, the road surface data collection sub-module 353 searches the map database 4 to acquire road surface data corresponding to the coordinates of the estimated position, searches the map database 4 according to the current position, and determines whether the vehicle is Acquire moving waypoint information and transmit road surface data to the route tracking sub-module 355 . In a communication operation A 19 , the vehicle dynamic state detection module 33 transmits the vehicle dynamic information to the route tracking sub-module 355 . The route tracking sub-module 355 generates a steering command according to the road surface data and vehicle dynamic information, and transmits the steering command to the steering angle control module 37 . The steering angle control module 37 determines the steering angle according to the steering command, and controls the vehicle body 5 according to the steering angle in communication operation A21.

Please refer to FIG. 13, which is a flowchart of a method for controlling an autonomous vehicle, according to one embodiment of the present disclosure. As shown in FIG. 13 , the method for controlling an autonomous vehicle includes step S31: estimating an estimated position after a processing time and a vehicle chassis control delay time according to the current position and current speed of the vehicle; step S32; : obtaining target road surface data corresponding to the estimated position from the map database; and step S33: generating a steering command according to the target road surface data to control the vehicle. The contents of steps S31-S33 are the same as those described in the embodiment of FIGS. 11 and 12, so they will not be described again.

With the above configuration, the method and system for extracting road surface data disclosed in the present disclosure classify the environment detection data to obtain the ground data, generate the ground model according to the ground data, and fill in the missing area. The ground model is optimized by using the optimized ground model, and the road surface attributes and road surface damage conditions are acquired, thereby obtaining highly complete road surface data. The method and system for extracting road surface data disclosed in the present disclosure can be applied to build and update high-definition maps to achieve high-definition maps, and can be used to control self-driving vehicles. It can be applied to achieve a high degree of agreement between the control strategy of an autonomous vehicle and the actual road surface conditions. The method and system for controlling an autonomous vehicle can control the vehicle according to the high-quality road surface data obtained by the method of extracting road surface data described above, and provide information on environmental displacement (road surface attributes and road surface damage). It solves the problem that the control strategy of the self-driving car cannot converge to the ideal route due to the lack of , and makes the vehicle travel route match the ideal route.

Although embodiments of the present disclosure are disclosed above, this is not intended to limit the spirit of the present disclosure. Modifications and changes that do not depart from the spirit and scope of this disclosure are included within the scope of this disclosure. For the scope of protection defined in this disclosure, refer to the appended claims.

1: System for extracting road surface data 11: Data input device 13: Storage device 15: Processing device 2: Environment detector z: Vertical axis PC: Point cloud VC: Voxel row OV: Original voxel TV: Target voxel group B1 : Area SL1: Cutting line C0: Original curved surface C1: First curved surface C2: Second curved surface D1: First distance D2: Second distance DR1: Damage area 3: System for controlling an automatic driving vehicle 31: Positioning module 33: Vehicle dynamic state detection module 35: Processing module 37: Steering angle control module 4: Map database 351: Moving distance estimation sub-module 353: Road surface data collection sub-module 355: Route tracking sub-module 5: Vehicle body A11-A21: Communication operation

TECHNICAL FIELD The present disclosure relates to methods for extracting road surface data, and more particularly to methods for extracting road surface data, including building a ground model.

With advances in artificial intelligence and related technologies, self-driving cars are being recognized as the future of transportation. The global automotive industry is investing in developing the field of autonomous driving. A self-driving car can program an appropriate driving route according to the current situation of the self-driving car and road conditions. Therefore, acquiring precise road surface data is one of the important issues in the field of autonomous driving.

With the rapid development of urban areas and the improvement of computer software technology, there is a growing demand for acquisition and analysis of road surface data not only in the field of autonomous driving, but also in areas such as planning and maintenance of urban redevelopment, car navigation, and location information services. increasing.

In view of the above description, the present disclosure provides methods and systems for acquiring road surface data and methods and systems for controlling self-driving vehicles.

According to one embodiment of the present disclosure, a method for obtaining road surface data includes executing the following steps in a processing device: obtaining ground data from first environment detection data in a detection area; obtaining a first ground model from the ground data; filling missing areas of the first ground model to obtain a second ground model; obtaining a plurality of road surface attributes of the detection area according to the second ground model; obtaining; performing road surface damage identification using the second environment detection data of the detection area and the second ground model; and storing the plurality of road surface attributes and road surface damage identification results as a plurality of road surface data of the detection area. to do.

According to one embodiment of the present disclosure, a method for controlling an autonomous vehicle includes executing the following steps in a processing device: current vehicle position, speed, processing delay time, and chassis control delay time. estimating the position of the vehicle upon activation of a subsequent chassis control command ; obtaining target road surface data corresponding to the estimated position from a map database; and generating a chassis control command according to the target road surface data. to control the vehicle. The map database comprises a plurality of pre- stored road surface data . The step of creating pre- stored road surface data includes obtaining ground data from the first environment detection data in the detection area; obtaining a first ground model from the ground data; obtaining a second ground model by compensating for the missing area of the second ground model; obtaining a plurality of road surface attributes of the detection area according to the second ground model; second environment detection data of the detection area and the second ground model; and storing the plurality of road surface attributes and the results of the road surface damage identification as a plurality of pre-stored road surface data in the detection area.

According to one embodiment of the present disclosure, a system for road surface data extraction includes a data input device, a storage device, and a processing device, the processing device being connected to the data input device and the storage device. The data input device is configured to receive first environmental sensing data and second environmental sensing data. The processing device is configured to perform the following steps: obtaining ground data from the first environment detection data; obtaining a first ground model using the ground data; obtaining the first ground model. obtaining a second ground model by compensating for the missing area of the second ground model; obtaining a plurality of road surface attributes of the detection area based on the second ground model; combining the second environment detection data with the second ground model performing road surface damage identification using a plurality of road surface attributes and road surface damage identification results as a plurality of road surface data in the detection area.

According to one embodiment of the present disclosure, a system for controlling an autonomous vehicle comprises a positioning module, a vehicle dynamic state sensing module, a chassis control module, and a processing module, the processing module comprising the positioning module, the vehicle It is connected to the dynamic condition sensing module and the chassis control module. The positioning module is configured to obtain the current position of the vehicle. The vehicle dynamic state sensing module is configured to obtain the current speed of the vehicle. The chassis control module receives chassis control commands and manipulates certain changes within the vehicle chassis, including steering, acceleration, braking and gearshifting . The processing module is configured to perform the following steps: using the vehicle position, velocity, computational delay time, and chassis control delay time to estimate the position of the vehicle upon activation of the next chassis control command; obtaining target road surface data corresponding to the estimated position from a map database; generating chassis control instructions according to the target road surface data to control the vehicle. The map database comprises a plurality of pre- stored road surface data . The road surface data extraction processing includes acquiring ground data from the first environment detection data in the detection area, acquiring the first ground model using the ground data, and filling the missing area of the first ground model. obtaining a second ground model using the second ground model; obtaining a plurality of road surface attributes of the detection area according to the second ground model; and identifying road damage using the second environment detection data of the detection area and the second ground model. and storing the plurality of road surface attributes and road surface damage identification results as a plurality of pre-stored road surface data in the detection area.

The above content and the following detailed description of the disclosure are used to illustrate the concept and spirit of the disclosure and to provide further explanation of the scope of the disclosure.

1 is a functional block diagram of an operating environment of a system for extracting road surface data according to one embodiment of the present disclosure; FIG. 4 is a flowchart of a method for extracting road surface data, according to one embodiment of the present disclosure; 4 is a flowchart of steps for obtaining ground surface data in a method for extracting road surface data, according to one embodiment of the present disclosure; FIG. 4 is a schematic diagram of performing the step of obtaining ground surface data in a method for extracting road surface data, according to one embodiment of the present disclosure; 4 is a flow chart of the steps of filling missing areas in a method for extracting road surface data, according to one embodiment of the present disclosure. FIG. 3 is a schematic diagram of environment detection data in a detection area, a first ground model, and a second ground model, according to one embodiment of the present disclosure; 4 is a flowchart of road damage identification steps in a method for extracting road surface data, according to one embodiment of the present disclosure. FIG. 4 is a schematic diagram of performing the step of road damage identification in a method for extracting road surface data, according to one embodiment of the present disclosure; FIG. 4 is a schematic diagram of performing the step of road damage identification in a method for extracting road surface data, according to one embodiment of the present disclosure; FIG. 4 is a schematic diagram of performing the step of road damage identification in a method for extracting road surface data, according to one embodiment of the present disclosure; 1 is a functional block diagram of an operating environment of a system for controlling an autonomous vehicle, according to one embodiment of the present disclosure; FIG. 1 is a schematic diagram of a method for controlling a self-driving vehicle performed by a system for controlling a self-driving vehicle, according to one embodiment of the present disclosure; FIG. 4 is a flowchart of a method for controlling an autonomous vehicle, according to one embodiment of the present disclosure;

In the following detailed description, for purposes of explanation, numerous specific details are set forth in order to provide a thorough understanding of the disclosed embodiments. The concepts and features of the present invention can be readily understood by those skilled in the art from the description, claims, and drawings disclosed herein. The following embodiments further illustrate various aspects of the invention, but are not intended to limit the scope of the invention.

Please refer to FIG. 1, which is a functional block diagram of the operating environment of a system for extracting road surface data, according to one embodiment of the present disclosure. The system 1 for extracting road surface data can obtain the environment detection data of a certain detection area, and consequently build a ground model of the detection area based on the detection data of the detection area. A system 1 for extracting road surface data performs interpolation operations on the ground model to optimize the ground model, and extracts multiple road surface attributes (e.g., bank angle , slope, radius of curvature, etc.) and road surface conditions (e.g., , potholes , ruts, etc.) can be obtained. As shown in FIG. 1, a system 1 for extracting road surface data comprises a data input device 11, a storage device 13 and a processing device 15, the processing device 15 being connected to the data input device 11 through a wired or wireless connection. It is connected to device 13 .

The data input device 11 may comprise a wireless communication module using communication technologies such as Wi-Fi, 4G, 5G. The data input device 11 is connected to the environment detector 2 and configured to receive environment sensing data detected by the environment detector 2 . Although FIG. 1 shows a typical example with one environment detector 2 , the data input device 11 may be connected to multiple environment detectors 2 . The environment detector 2 may be a LIDAR, and environment detection data to be detected may be a point cloud. The environment detector 2 may be fixed or mounted on a moving platform such as an aircraft, unmanned aerial vehicle (UAV), or vehicle. Here, the vehicle is, for example, a radar plot vehicle, an automatic driving vehicle, or the like. The data input device 11 provides the environment detection data acquired by one or more environment detectors 2 to the processing device 15 , and the processing device 15 performs the extraction of the road surface data of the particular detection area. The particular detection area depends on the detection range of the one or more environmental detectors 2 .

Storage device 13 may comprise, but is not limited to, flash memory, hard disk drive (HDD), solid state disk (SSD), dynamic random access memory (DRAM), static random access memory (SRAM). The storage device 13 can store road surface data extracted by the processing device 15 as described above. In one embodiment, storage device 13, data input device 11, and processing device 15 are components of the same server. In other embodiments, storage device 13 and processing device 15 are provided in different servers and are connected to each other by wired or wireless connections.

Processing unit 15 is a single processor or collection of multiple processors, such as, but not limited to, a central processing unit (CPU) or graphics processing unit (GPU). The processing device 15 acquires environment detection data of the detection area from the data input device 11, processes the environment detection data to extract road surface data of the detection area, and stores or stores the extracted road surface data in the storage device 13. It is arranged to update the road surface data stored in the device 13 . The road surface data may include road surface attributes and the results of road surface damage identification, and performing methods for extracting the road surface data are described below.

See FIGS. 1 and 2. FIG. 2, which is a flowchart of a method for extracting road surface data according to one embodiment of the present disclosure. As shown in FIG. 2, a method for extracting road surface data includes step S11: obtaining ground surface data from the first environment detection data in the detection area; Obtaining a model, Step S13: Filling in missing areas of the first ground model to obtain a second ground model, Step S14: Obtaining a plurality of attributes of the detection area according to the second ground model. Step S15: performing road surface damage identification using the second environment detection data of the detection region and the second ground model; Including storing as data. It should be mentioned that the present disclosure does not limit the order of steps S14 and S15. The method for extracting road surface data shown in FIG. 2 can, but is not limited to, be performed by the processing device 15 as shown by the system 1 for extracting road surface data. For ease of understanding, the steps of the method for extracting road surface data will be described using the operation by the processing device 15 as an example.

In step S11, the processing device 15 can acquire the first environment detection data of the detection area from the environment detector 2 via the data input device 11, and acquire the ground data from the first environment detection data. More specifically, see FIGS. 3 and 4. FIG . As shown in FIG. 3, the step of obtaining the ground data includes sub-step S111: dividing the first environment detection data (point cloud) into a plurality of original voxels , dividing the original voxels sharing a common horizontal position into a plurality of voxels; grouping into voxel columns, sub-step S112: obtaining at most one target voxel group from each original voxel group, where each target voxel group has a substantially upward surface normal and a predetermined the lowest vertically contiguous group of non-empty original voxels with a vertical diffusion parameter less than the value of substep S113: For each target voxel group, calculate a plurality of height values by calculating the mean of the point heights and sub-step S114: generating ground data using the plurality of altitude values.

Each of the substeps is further described. In substeps S111 and S113, the vertical axis may be the axis perpendicular to the horizontal plane. Sub-step S112 may include obtaining a plurality of vertically consecutive voxels from each voxel column and recognizing the lowest voxel that satisfies a plurality of specific conditions as the target voxel. The specific conditions are that (1) the normal of the surface fitted to the points contained in the vertically contiguous group of voxels is nearly upward , and (2) the normal of the points contained in the vertically contiguous group of voxels It may include that the diffusion coefficient is less than or equal to a predetermined value. Among voxel groups that satisfy specific conditions (1) and (2) and exist in the same voxel row, the lowest voxel group is the voxel group having the lowest altitude. The surface normal (surface normal) in condition (1) may be determined by principal component analysis and scanning trajectory. For example, the scan trajectory may be the scan trajectory of the environment detector that generated the environment detection data. Specifically, the principal component analysis may estimate two normal candidates in opposite directions, where the normal candidate lying on the same side of the scan trajectory is selected as the surface normal. Whether the surface normal is nearly upward is determined by checking that the angle between the surface normal and the vertical axis is less than a threshold. The vertical diffusion parameter in condition (2) is the height difference between the highest and lowest points in the voxels. For example, in one embodiment, processor 15 may eliminate all voxel groups having a vertical diffusion parameter greater than 30 cm.

Sub-step S114 performs a region growing operation or a connecting quantification operation on a plurality of altitude values obtained from a plurality of target voxel groups to form a plurality of altitude value connecting groups, and a scanning trajectory and one having overlap. Selecting one or more altitude value connections as ground data may be included.

Take the point cloud PC shown in FIG. 4 as an example. A voxel column VC consists of the original voxels OV arranged along the z-axis. Note that although FIG. 4 shows only one voxel train, in practice all point clouds PC can be divided into a plurality of voxel trains. The number of voxel columns VC and the number of original voxels OV that make up each row are not limited to the present disclosure. In FIG. 4, the voxel column VC contains eight original voxels OV labeled voxels 1-8 from bottom to top. Voxels 1 and 4-7 are non-empty since they contain points. Voxels 2, 3 and 8 are empty since they contain no points. Voxel 1 is itself a group of vertically contiguous non-empty original voxels ; It is recognized as a voxel group TV. Voxels 4-7 form another vertically contiguous group of non-empty original voxels, but since the points have a wide vertical diffusion parameter, they are not recognized as a target voxel group. The mean height of the points in the target voxels TV serves as the mean height of the target voxels TV and is used to generate the ground data. The method of generating the ground data is as described above and will not be repeated here.

Please refer to FIGS. 1 and 2 again. At step S12, the processing device 15 uses the ground data to obtain a first ground model. In one embodiment, the processor 15 directly uses the ground data obtained in step S11 as the first ground model. In another embodiment, processor 15 fits a curved surface model to the ground data to obtain a first ground model. In step S13, the processing device 15 fills in the missing area in the first ground model to acquire the second ground model . As shown in FIG. 5, the step for filling the missing area includes sub-step S131 : transforming the first ground model into a height map, a two-dimensional image whose pixels are altitude values from the first ground model; substep S132: performing digital inpainting on the height map to generate an inpainted height map ; and substep S133: transforming the inpainted height map into a three-dimensional model as a second ground model. can include A height map may be, for example, a regular grid. The missing area may be a portion of the ground that qualifies as an obscured view of the environment detector. Methods for digital restoration are generally capable of filling defective regions of arbitrary shape. An example of a method for digital inpainting solves a function f: Ω→R that minimizes the whole over Ω of the Laplacian function (Δf) ² , where Ω is the subset of the two-dimensional plane corresponding to the defect region, At the same time, it is necessary to satisfy the continuity of the function on the boundary ∂Ω of the missing region.

Please refer to FIG. 6, which is a schematic diagram of the environment detection data of the detection area, the first ground model, and the second ground model, according to one embodiment of the present disclosure. FIG. 6 shows the environment detection data (point cloud) of the detection areas A to C. Here, the first ground model uses the above step S11 and step S12 in FIG. , and the second ground model is obtained by performing step S13 in FIG. 2 on the first ground model corresponding to the detection area AC.

Please refer to FIGS. 1 and 2 again. In step S14, the processing device 15 acquires road surface attributes of the detection area according to the second ground model. More specifically, the processing unit 15 obtains from the second ground model a plurality of altitude values within an area formed by a predetermined driving route that extends a predetermined distance outward, and the (road surface) of the predetermined driving route. It can calculate road surface attributes such as slope, slope and radius of curvature of the surrounding terrain. There may be one or more predetermined driving routes. For example, the predetermined driving route may be a Google Maps suggested route or a pre-programmed driving route by the user, and the predetermined distance may be a typical paved road or the width of the vehicle, but the predetermined driving route and the predetermined distance are not limited to these.

In step S15, the processing device 15 performs road damage identification using the second environment detection data of the detection area and the second road surface model. Points belonging to certain road surface damage (e.g. potholes or ruts) will be excluded from the road surface data output by step S11 (because they are no longer part of a larger road surface and contiguous configuration). , leaving missing areas in the road surface data, which are later filled in in step S13 to generate a second road surface model without road damage. Using this second road surface model as a reference, processor 15 searches for differences from the model from the second environment sensing data and flags any such differences as possible road damage. . In one embodiment, the second environment detection data and the first environment detection data for generating the second road surface model are the same data (detecting road damage from one environment detection session). In another embodiment, the second environmental sensing data and the first environmental sensing data belong to the same sensing region but are sensed at different times . In this embodiment, since new road surface damage occurs and old damage is repaired, the processing device 15 constructs the second road surface model and acquires the second environment detection data after a certain period of time has elapsed. , which allows the processor 15 to track changes in the location and characteristics of road damage. In particular, the processing device may periodically acquire environment detection data detected by the environment detector 2 and periodically update the road surface data.

Step S15 will be further described. Please refer to FIG. 7, which is a flowchart of road damage identification steps in a method for extracting road surface data, according to one embodiment of the present disclosure. As described above, the first road surface model may be the road surface data obtained in step S11, or may be a curved surface model fitted to the road surface data. Obtained by supplementing . Therefore, the second road surface model can have a curved surface, hereinafter referred to as the original curved surface. As shown in FIG. 7, the steps of road damage identification include substep S151: shifting the original curved surface down a first distance along the vertical axis to create a first curved surface; down a second distance that is greater than the first distance along the vertical axis to create a second curved surface; and substep S153: below the first curved surface, and , identifying any points above the second curved surface and flagging these points as potential road damage points. In particular, the magnitude of the first distance may relate to the accuracy of the point cloud, which depends on the accuracy of the environmental detector scan; may be environment dependent errors such as reflections from a wet road surface that cause the actual road surface to appear far below. For example, if the height of false points due to reflections from a wet road surface is typically 6 cm below the actual road surface, then the second distance may be set to 6 cm. Note that the order of sub-step S151 and sub-step S152 is not limited in the present disclosure.

The sub-steps described above will be explained with schematic diagrams. Please refer to FIGS. 8-10, which are schematic diagrams of performing the steps of road damage identification in a method for extracting road surface data, according to one embodiment of the present disclosure. FIG. 8(a) exemplarily shows the second environment detection data, and an enlarged view of the area B1 is shown in FIG. 8(b). FIG. 9 exemplarily shows end views of the second environment detection data, the original curved surface C0, the first curved surface C1, and the second curved surface C2 cut by the cutting line SL1. A first surface C1 is obtained by shifting the original surface C1 along the vertical axis z down a first distance D1, and a second surface C2 is obtained by shifting the original surface C0 along the vertical axis z by a second distance D1. 2 by shifting down a distance D2. In this embodiment, the first distance D1 is 3 cm and the second distance D1 is 6 cm, but the disclosure is not so limited. As shown in FIG. 10(b), points between the first curve C1 and the second curve C2 in the second environment detection data (point group) are determined as points forming the damaged region DR1. . FIGS. 10(a) and 10(b) show data schematic diagrams of marked road damage locations in the second environment detection data shown in FIGS. 8(a) and 8(b), respectively.

Please refer to FIGS. 1 and 2 again. In step S16, the processing device 15 stores the road surface attribute and the road surface damage identification result as the road surface data of the detection area in the storage device. Specifically, road surface attributes and road damage identification results can be stored in one of several possible formats, including irregular triangulation, regular grid, and waypoints. The waypoints may be a plurality of points on the predetermined driving path, each of the points on the predetermined driving path having road surface attributes and/or road damage identification results (e.g., damaged/undamaged) at the corresponding location. ). In one embodiment, the road surface data may further include a second road surface model, which may be stored in one of the above formats along with the road surface attributes and road damage identification results. In particular, data such as road surface model, road surface attributes, and road damage identification results can be stored in one of the following formats: irregular triangulation, regular grid, and waypoints. One or more of the above data may be stored as the same or different data, and the disclosure is not so limited.

In particular, the method for extracting road surface data and the system for extracting road surface data described in the above one or more embodiments apply to building high-definition maps and/or controlling self-driving vehicles. be able to.

Please refer to FIG. 11, which is a functional block diagram of the operating environment of a system for controlling an autonomous vehicle, according to one embodiment of the present disclosure. As shown in FIG. 11 , the system for controlling the autonomous vehicle 3 comprises a positioning module 31 , a vehicle dynamic state detection module 33 , a processing module 35 and a chassis control module 37 . The processing module 35 may be wired or wirelessly connected to the positioning module 31 , the vehicle dynamic state sensing module 33 and the chassis control module 37 , and the positioning module 31 and the processing module 35 are wired or wirelessly connected to the map database 4 . may be connected. The map database 4 includes maps of one or more detection areas each containing pre-stored road surface data. The details are not described again, as they can be obtained by the method for and the system for extracting the road surface data. In particular, the map database 4 can be realized by the storage device in the system for extracting road surface data described in the above embodiments.

The positioning module 31 is configured to obtain the current position of the vehicle to be controlled and may include, but is not limited to, position sensors or the like. The vehicle dynamic state sensing module 33 is configured to obtain the current speed of the vehicle and may comprise dynamic sensors such as wheel speed sensor, steering wheel angle sensor, yaw angle sensor, deceleration sensor, etc., which include: Not limited. The chassis control module 37 is configured to control the vehicle chassis in accordance with the chassis control instructions and includes elements such as throttle bodies, solenoid valves, relays, etc. that control braking and driving forces of the vehicle to control yaw motion of the vehicle. It can be provided, but is not limited to these. The processing module 35 may be, for example, a microcontroller within an electronic control unit (ECU) or a central processing unit within a microcontroller , and may be used to determine the position of the vehicle during the actuation of chassis control commands (e.g., to calculate steering commands). and the time between when the steering command is issued and when the wheel actually turns), acquires the target road surface data corresponding to the estimated position from the map database 4, and acquires the acquired target It is configured to generate chassis control instructions according to the road surface data to control the vehicle via the chassis control module 37 . The estimated position is contained in one of the one or more detection regions. The processing module 35 can search the map database 4 for road surface data of the detection area where the estimated position exists, and obtain target road surface data of the estimated position from the road surface data .

The operation of the system for controlling the autonomous vehicle 3 is further described. Please refer to FIG. 12, which is a schematic diagram of a system for controlling an autonomous vehicle, according to one embodiment of the present disclosure. The processing module 35 may comprise a displacement vector estimation sub-module 351 , a road surface data collection sub-module 353 and a route tracking sub-module 355 . For example, the above sub-modules may be implemented by one or more instructions or programs each having functions of estimating a displacement vector , collecting road surface data, and tracking a route.

In communication operations A 11 and A 12 , the positioning module 31 obtains the coordinates of the current position of the vehicle on the map and provides the coordinates of the current position to the road surface data collection sub-module 353 . In communication operations A13 and A14, the vehicle dynamic state sensing module 33 detects vehicle dynamic information (such as current speed, current yaw angular velocity, etc.) from the vehicle body 5, and transmits the vehicle dynamic information to the displacement vector estimation sub-module. provided to module 351; The displacement vector estimation sub-module 351 adds the processing delay time of the processing module 35 and the vehicle chassis control delay time of the vehicle body 5 to obtain the total delay time, and multiplies the total delay time by the current speed to estimate Obtain the displacement vector and transmit the estimated displacement vector to the road surface data collection sub-module 353 in communication operation A15. For example, the processing delay time of processing module 35 is the total estimated time to perform data processing by displacement vector estimation sub-module 351 , road surface data collection sub-module 353 , and route tracking sub-module 355 . The vehicle chassis delay time may then be the time between the time the steering command is received and the time the wheel actually turns. The road surface data collection sub-module 353 obtains the coordinates of the estimated position using the current position and the estimated displacement vector . In communication operations A16-A18, the track track sub-module 355 retrieves the map database 4 to search the road surface data corresponding to the coordinates of the estimated position (the road surface data instructs the vehicle on how to proceed). ) , and transmit the road surface data to the route tracking sub-module 355 . In a communication operation A 19 , the vehicle dynamic state detection module 33 transmits the vehicle dynamic information to the route tracking sub-module 355 . The path tracking sub-module 355 generates chassis control instructions according to the road surface data and vehicle dynamics , and transmits the chassis control instructions to the chassis control module 37 . The chassis control module 37 controls the vehicle body 5 according to the chassis control command in communication operation A21.

Please refer to FIG. 13, which is a flowchart of a method for controlling an autonomous vehicle, according to one embodiment of the present disclosure. As shown in FIG. 13, the method for controlling an autonomous vehicle includes step S31: using information about the chassis control delay time and the current position and speed of the vehicle to generate chassis control commands (e.g., steering, gas pedal, braking, gear shift); ), step S32: acquiring target road surface data corresponding to the estimated position from the map database, step S33: generating a chassis control command according to the target road surface data, and may include controlling the The contents of steps S31-S33 are the same as those described in the embodiment of FIGS. 11 and 12, so they will not be described again.

With the above configuration, the method and system for extracting road surface data disclosed in the present disclosure classify the environment detection data to obtain the ground data, generate the ground model according to the ground data, and fill in the missing area. This optimizes the ground model , acquires road surface attributes and road surface damage conditions , and enables acquisition of highly complete road surface data. The methods and systems for extracting road surface data disclosed in this disclosure can be applied to build and update high-precision maps and can be applied to control self-driving vehicles, thereby , can provide a more precise and accurate perception of surrounding road conditions for control decisions. The method and system for controlling an autonomous vehicle can control the vehicle according to the high degree of road surface data obtained by the above-described method of extracting road surface data , which enables (bank angle, slope, road damage etc.), it is possible to solve the problem that the self-driving car cannot faithfully follow the ideal route due to the lack of recognition of the environmental situation.

Although embodiments of the present disclosure are disclosed above, this is not intended to limit the spirit of the present disclosure. Modifications and changes that do not depart from the spirit and scope of this disclosure are included within the scope of this disclosure. For the scope of protection defined in this disclosure, refer to the appended claims.

1: System for extracting road surface data 11: Data input device 13: Storage device 15: Processing device 2: Environment detector z: Vertical axis PC: Point cloud VC: Voxel row OV: Original voxel TV: Target voxel group B1 : Area SL1: Cutting line C0: Original curved surface C1: First curved surface C2: Second curved surface D1: First distance D2: Second distance DR1: Damage area 3: System for controlling an automatic driving vehicle 31: Positioning module 33: Vehicle dynamic state detection module 35: Processing module 37: Chassis control module 4: Map database 351: Displacement vector estimation sub-module 353: Road surface data collection sub-module 355: Route tracking sub-module 5: Vehicle body A11 -A21: Communication operation

Claims (22)

A method for extracting road surface data including the following steps executed by a processing device, said steps:
obtaining ground data from the first environmental sensing data in the sensing area;
obtaining a first ground model using the ground data;
Acquiring a second ground model by compensating for the missing area of the first ground model;
obtaining a plurality of road surface attributes of the detection area according to the second ground model;
performing road surface damage identification using the second environment detection data of the detection area and the second ground model; and using the plurality of road surface attributes and the results of the road surface damage identification as a plurality of road surface data of the detection area. A method for extracting road surface data, including storing. The first environment detection data is a point cloud, and acquiring the ground data from the first environment detection data of the detection area includes:
dividing the point cloud into a plurality of original voxels;
obtaining a plurality of target voxels from the plurality of original voxels;
calculating a plurality of elevation values for the plurality of target voxels; and using the plurality of elevation values to generate the ground data;
the plurality of original voxels forming a plurality of voxel columns, each of the plurality of voxel columns extending along a vertical axis;
Each of the plurality of target voxel groups is included in a plurality of voxel columns, a surface normal of a plurality of points in each of the plurality of target voxel groups is upward, and a vertical diffusion parameter of the plurality of points is equal to or less than a predetermined value. and
2. The method for extracting road surface data of claim 1, wherein each of said plurality of altitude values represents an average value of values on said vertical axis of said plurality of points of said target voxels. 3. The method for extracting road surface data according to claim 2, wherein the surface normal is determined by principal component analysis and scanning trajectories. Generating the ground data using the plurality of altitude values includes:
performing a region expanding operation or a connection volume operation on the plurality of altitude values to create a plurality of connected groups of altitude values; and according to overlapping regions of each of the plurality of connected groups of altitude values and the scan trajectory. 3. A method for extracting road surface data according to claim 2, comprising selecting at least one of said plurality of connected groups of elevation values as ground surface data. Acquiring a second ground model by compensating for the missing area of the first ground model,
converting the first ground model into two-dimensional data comprising a plurality of storage units each storing a plurality of altitude values of the first ground model;
performing digital inpainting on the two-dimensional data to generate inpainted two-dimensional data; and transforming the inpainted two-dimensional data into a three-dimensional model as the second ground model. A method for extracting road surface data according to claim 1. The second environment detection data is a point cloud, the second ground model includes an original curved surface, and the road damage identification is performed using the second ground model of the detection area and the second ground model. to do
shifting the original curved surface down a first distance along a vertical axis to generate a first curved surface;
shifting the original curved surface down a second distance greater than the first distance along the vertical axis to generate a second curved surface; and 2. The method for extracting road surface data of claim 1, comprising determining a plurality of points between the second curved surface to form one or more damaged regions. 7. For extracting road surface data according to claim 6, wherein said first distance is related to the accuracy of said point cloud and said second distance is related to detection error of an environment detector generating said point cloud. the method of. 2. The method for extracting road surface data of claim 1, wherein the plurality of road surface attributes are stored in one of a format of irregular triangulation, regular grid, and waypoints. 2. The method for extracting road surface data according to claim 1, wherein said first environmental detection data and said second environmental detection data are detected at different times. A method for controlling a self-driving vehicle comprising the following steps performed by a processing unit, said steps:
estimating an estimated position after a processing time and a vehicle chassis control delay time according to the current position and current velocity of the vehicle;
obtaining target road surface data corresponding to the estimated position from a map database; and generating steering commands according to the target road surface data to control the vehicle;
The map database comprises a plurality of pre-stored road surface data of a detection area, the estimated position is located within the detection area, and the target road surface data is included in the plurality of pre-stored road surface data. , the plurality of road surface data stored in advance are acquired by a road surface data extraction process, and the road surface data extraction process includes:
obtaining ground data from the first environmental detection data of the detection area;
obtaining a first ground model from the ground data;
Acquiring a second ground model by compensating for the missing area of the first ground model;
obtaining a plurality of road surface attributes of the detection area according to the second ground model;
performing road surface damage identification using the second environment detection data of the detection area and the second ground model; A method for controlling a self-driving vehicle, including storing as road surface data. estimating the estimated position after the processing time and the vehicle chassis control delay time according to the current position and the current velocity of the vehicle;
summing the processor processing time and the vehicle chassis control delay time to obtain an estimate of the total delay time;
multiplying the total delay time and the current velocity to obtain an estimated distance traveled; and using the current position and the estimated distance traveled to obtain the estimated position. A method for controlling an autonomous vehicle according to claim 10. a data input device configured to receive first environmental sensing data and second environmental sensing data;
Storage device,
a processing device connected to the data input device and the pre-storage device and configured to perform the steps of:
obtaining ground data from the first environment detection data;
obtaining a first ground model using the ground data;
Acquiring a second ground model by compensating for the missing area of the first ground model;
obtaining a plurality of road surface attributes of a detection area according to the second ground model;
performing road surface damage identification using the second environment detection data and the second ground model; and storing the plurality of road surface attributes and results of the road surface damage identification as a plurality of road surface data of the detection area. A system for extracting road surface data, including the first environment detection data is a point cloud;
The processing device divides the point cloud into a plurality of original voxels to obtain a plurality of target voxels, calculates a plurality of altitude values of the plurality of target voxels, and uses the plurality of altitude values to determine the ground level. configured to generate data,
The plurality of original voxels form a plurality of voxel columns each extending along a vertical axis, the plurality of target voxel groups are included in the plurality of voxel columns, and the plurality of target voxels in each of the plurality of target voxel columns. the surface normal of a point is upward, the vertical diffusion parameter of the plurality of points is less than or equal to a predetermined value, and each of the plurality of altitude values is along the vertical axis of the plurality of points within the plurality of target voxels. 13. A system for extracting road surface data as claimed in claim 12, representing an average of the above values. 14. The system for extracting road surface data of claim 13, wherein the surface normal is determined by principal component analysis and scan trajectory. The processing device executes a region expansion operation or a connection amount operation on the plurality of altitude values to create a plurality of connected groups of altitude values, and each of the plurality of connected groups of altitude values and a scan trajectory. 14. The system for extracting road surface data according to claim 13, configured to select at least one of said plurality of connected groups of altitude values as said ground surface data according to an amount of overlap. The processing unit is configured to convert the first ground model to two-dimensional data, perform digital inpainting on the two-dimensional data to generate inpainted two-dimensional data, and converting the two-dimensional data into a three-dimensional model as a second ground model;
13. The system for extracting road surface data according to claim 12, wherein said two-dimensional data comprises a plurality of storage units each storing a plurality of altitude values of said first ground model. The second environment detection data is a point cloud, the second ground model includes an original curved surface, and the processing unit shifts the original curve down a first distance along a vertical axis. creating a first curved surface, shifting the original curved surface down a second distance along the vertical axis that is greater than the first distance to create a second curved surface, and creating the 13. The system for extracting road surface data of claim 12, wherein a plurality of points between the first curved surface and the second curved surface are determined to create one or more damaged areas. 18. The method of extracting road surface data of claim 17, wherein the first distance relates to the accuracy of the point cloud and the second distance relates to error due to environmental factors generated by the point cloud. system for. 13. The system for extracting road surface data of claim 12, wherein the plurality of road surface attributes are stored in one of a format of irregular triangulation, regular grid, and waypoints. 13. The system for extracting road surface data as recited in claim 12, wherein said first environmental detection data and said second environmental detection data are detected at different times. A system for controlling an autonomous vehicle,
a positioning module configured to obtain the current position of the vehicle;
a vehicle dynamic state sensing module configured to obtain a current speed of the vehicle;
a steering angle control module configured to control the vehicle according to steering commands;
a processing module connected to said positioning module, said vehicle dynamic state sensing module and said steering angle control module and configured to perform the following steps, said steps comprising:
estimating an estimated position after a processing time and a vehicle chassis control delay time according to the current position of the vehicle and the current speed of the vehicle;
obtaining target road surface data corresponding to the estimated position from a map database; and generating steering commands according to the target road surface data to control the vehicle;
The map database comprises a plurality of pre-stored road surface data of a detection area, the estimated position is within the detection area, and the target road surface data is included in the plurality of pre-stored road surface data. , the plurality of road surface data stored in advance are acquired by a road surface data extraction process, and the road surface data extraction process includes:
obtaining ground data from the first environmental sensing data in the sensing area;
obtaining a first ground model using the ground data;
Acquiring a second ground model by compensating for the missing area of the first ground model;
obtaining a plurality of road surface attributes of the detection area according to the second ground model;
performing road surface damage identification using the second environment detection data of the detection area and the second ground model; A system for controlling self-driving cars, including storing as multiple road surface data. The processing module adds the processing time of the processing module and the vehicle chassis control delay time to obtain a total delay time, and multiplies the total delay time by the current speed to obtain an estimated traveled distance. 22. The system for controlling a self-driving vehicle according to claim 21, wherein an estimated position is obtained using the current position and the estimated distance traveled.

Images