Road model generation method and device, automobile and storage medium

Abstract

The application relates to a road model generation method and device, a vehicle and a storage medium. The method comprises the following steps: acquiring a first road model generated in the cloud, wherein the first road model at least comprises first road geometric information and first road topological information; acquiring vehicle-sensed map information detected at a vehicle end; matching the vehicle-sensed map information with the first road geometric information and the first road topological information of the first road model to obtain target road geometric information and target road topological information; and generating a target road model at the vehicle end according to the target road geometric information and the target road topological information. The scheme provided in the application improves the accuracy and stability of the road geometric structure at the vehicle end, and improves the efficiency of constructing the target road model by consuming the prior information of the first road model in the cloud.

Description

Technical Field

[0001] The present application relates to the field of autonomous driving technology, and in particular to a road model generation method, device, vehicle, and storage medium. Background Art

[0002] With the continuous development of autonomous driving technology, road models are playing an increasingly important role in vehicle navigation and control. Compared to ordinary high-precision maps, road models not only contain detailed road data such as road information, traffic signs, lane markings, and traffic lights, but also include road topology information that supports user driving decision-making.

[0003] In actual application scenarios, the cloud and vehicle-side independently construct road structures. The cloud leverages its beyond-line-of-sight (BVR) advantage to generate the cloud-side road structure, while the vehicle-side generates the vehicle-side road structure based on vehicle-side perception information. Related technologies require first matching the vehicle-side road structure with the cloud-side road structure. For successfully matched lanes, the cloud-side prior information is then applied to the vehicle-side lanes to generate a vehicle-side road model. However, due to the significant limitations and instability of real-time vehicle-side perception information, the road structure constructed in real time based on the vehicle is unstable and fluctuates, making it difficult to geometrically match the cloud-side road structure, and thus, unable to achieve a stable and effective match between the vehicle and cloud. Summary of the Invention

[0004] In order to solve or partially solve the problems existing in the related art, the present application provides a road model generation method, device, automobile and storage medium, which can improve the accuracy and stability of the generated road model.

[0005] A first aspect of the present application provides a road model generation method, comprising:

[0006] Obtaining a first road model generated in the cloud, where the first road model at least includes: first road geometry information and first road topology information;

[0007] Obtain vehicle-sensing map information detected by the vehicle side;

[0008] Matching the vehicle-sensing map information with first road geometry information and first road topology information of the first road model to obtain target road geometry information and target road topology information;

[0009] Generate the vehicle-side target road model based on the target road geometry information and target road topology information.

[0010] Optionally, matching the vehicle-sensing map information with the first road geometry information and the first road topology information of the first road model to obtain target road geometry information and target road topology information includes:

[0011] Matching the vehicle sensing map information with the first road geometry information to obtain target road geometry information;

[0012] Target road topology information is constructed based on the first road topology information, the vehicle sensing map information, and the target road geometry information.

[0013] Optionally, the first road topology information includes at least one or more of lane turning information, preceding and succeeding information, lane type, and user driving trajectory.

[0014] Optionally, matching the vehicle sensing map information with the first road geometry information to obtain target road geometry information includes:

[0015] Obtaining second road geometry information based on vehicle sensing map information;

[0016] The first road geometric information and the second road geometric information are geometrically matched, and target road geometric information is generated according to the matching result.

[0017] Optionally, geometrically matching the first road geometric information and the second road geometric information includes:

[0018] Determining a reliable range of ego vehicle perception, and determining target road geometry information based on the second road geometry information within the reliable range of ego vehicle perception;

[0019] When the vehicle exceeds the reliable perception range, the second road geometry information is corrected according to the first road geometry information to determine the target road geometry information.

[0020] Optionally, when the vehicle exceeds the reliable perception range, the second road geometry information is corrected according to the first road geometry information to determine the target road geometry information, including:

[0021] Obtaining, according to a preset geometric information conversion format, a first road centerline of the first road geometric information and a second road centerline of the second road geometric information;

[0022] The first road centerline and the second road centerline that are beyond the reliable perception range of the vehicle are weightedly fitted to determine the target road geometry information.

[0023] Optionally, constructing target road topology information according to the first road topology information, the vehicle sensing map information, and the target road geometry information includes:

[0024] Verifying the first road topology information, and correcting the first road topology information that fails to pass the verification according to the vehicle-sensing map information to obtain second road topology information;

[0025] Comparing the vehicle-sensing map information with the first road topology information to determine newly added road topology information;

[0026] Target road topology information is generated according to the newly added road topology information and the second road topology information.

[0027] Optionally, verifying the first road topology information includes:

[0028] Determining the rationality of the connection between the preceding and succeeding information of the first road topology information according to a preset judgment rule; and / or;

[0029] The accuracy of the lane turning information, the preceding and following information, and the lane type of the first road topology information is verified based on the vehicle sensing map information.

[0030] Optionally, comparing the vehicle-sensing map information and the first road topology information to determine the newly added road topology information includes:

[0031] Obtaining the driving trajectory and driving habits of the newly added user based on the vehicle sensing map information;

[0032] The newly added road topology information is generated according to the newly added user's driving trajectory and the newly added user's driving habits.

[0033] A second aspect of the present application provides a road model generation device, comprising:

[0034] A first acquisition module is configured to acquire a first road model generated in the cloud, where the first road model includes at least first road geometry information and first road topology information;

[0035] The second acquisition module is used to obtain vehicle-sensing map information detected by the vehicle side;

[0036] a processing module, configured to match the vehicle-sensing map information with first road geometry information and first road topology information of the first road model to obtain target road geometry information and target road topology information;

[0037] The determination module generates a vehicle-side target road model based on the target road geometry information and target road topology information.

[0038] A third aspect of the present application provides a vehicle, comprising:

[0039] processor; and

[0040] The memory stores executable codes thereon, and when the executable codes are executed by the processor, the processor is caused to execute the method described above.

[0041] A fourth aspect of the present application provides a computer-readable storage medium having executable code stored thereon. When the executable code is executed by a processor of an electronic device, the processor is caused to execute the method described above.

[0042] The technical solution provided by this application may have the following beneficial effects:

[0043] First, this application acquires a cloud-generated first road model and vehicle-based vehicle-sensing map information. Based on the a priori information of the first road model, the vehicle-based map information is integrated to generate a target road model on the vehicle. Compared to traditional methods of generating vehicle-based road models, this application incorporates prior information from the cloud into the vehicle-based road structure, saving computing power on the vehicle side to build the road model and improving the efficiency of generating the target road model.

[0044] Secondly, this application preserves the cloud's advantage in building road models beyond visual range by consuming the cloud's first road geometry and first road topology information. By matching vehicle-sensing map information, first road geometry, and first road topology information, the instability of road structures caused by the high volatility of vehicle-side perception information can be reduced, thereby building a more stable target road model. Furthermore, vehicle-sensing map information can collect new user driving trajectories in real time. By integrating these new user driving trajectories into the target road model, the target road model can update the user's driving characteristics in real time.

[0045] It should be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] The above and other objects, features and advantages of the present application will become more apparent by describing in more detail exemplary embodiments of the present application in conjunction with the accompanying drawings, wherein like reference numerals generally represent like components in the exemplary embodiments of the present application.

[0047] Figure 1 is a flowchart of a road model generation method shown in an embodiment of the present application;

[0048] Figure 2 is a schematic diagram of a first road model in the cloud shown in an embodiment of the present application;

[0049] Figure 3 is a schematic diagram of first road geometric information of a first road model shown in an embodiment of the present application;

[0050] Figure 4 is a schematic diagram of vehicle-sensing map information shown in an embodiment of the present application;

[0051] Figure 5 is a schematic diagram of a target road model shown in an embodiment of the present application;

[0052] Figure 6 Schematic diagram of the structure of the road model generation method and device shown in an embodiment of the present application;

[0053] Figure 7 It is a schematic structural diagram of a car shown in an embodiment of the present application. DETAILED DESCRIPTION

[0054] The following describes embodiments of the present application in more detail with reference to the accompanying drawings. Although the accompanying drawings illustrate embodiments of the present application, it should be understood that the present application can be implemented in various forms and should not be limited by the embodiments described herein. Rather, these embodiments are provided to make the present application more thorough and complete, and to fully convey the scope of the present application to those skilled in the art.

[0055] The terms used in this application are for the purpose of describing specific embodiments only and are not intended to limit this application. As used in this application and the appended claims, the singular forms "a," "an," "the," and "the" are intended to include the plural forms, unless the context clearly indicates otherwise. It should also be understood that the term "and / or" as used herein refers to and encompasses any and all possible combinations of one or more of the associated listed items.

[0056] It should be understood that although the terms "first", "second", "third", etc. may be used in this application to describe various information, this information should not be limited to these terms. These terms are only used to distinguish information of the same type from each other. For example, without departing from the scope of this application, the first information may also be referred to as the second information, and similarly, the second information may also be referred to as the first information. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the features. In the description of this application, the meaning of "plurality" is two or more, unless otherwise clearly and specifically defined.

[0057] With the continuous development of autonomous driving technology, road models are playing an increasingly important role in vehicle navigation and control. Compared to ordinary high-precision maps, road models not only contain detailed road data such as road information, traffic signs, lane markings, and traffic lights, but also include road topology information that supports user driving decision-making.

[0058] In actual application scenarios, the cloud leverages its beyond-line-of-sight (BVR) advantage to generate a cloud-based road model, while the vehicle generates the vehicle-based road geometry based on vehicle-based perception information. Related technologies match the vehicle-based road geometry to the cloud-based road structure. For successfully matched lanes, the cloud-based prior information is applied to the vehicle-based road model for that lane. However, due to the significant limitations and instability of real-time vehicle-based perception information, the road structure constructed in real time by the vehicle is unstable and fluctuates, making it difficult to geometrically match the cloud-based road structure, preventing a stable and effective match between the vehicle and cloud.

[0059] The present application provides a road model generation method, device, vehicle and storage medium, which can improve the accuracy and stability of the generated road model.

[0060] The technical solutions of the embodiments of the present application are described in detail below with reference to the accompanying drawings.

[0061] Figure 1 It is a flowchart of the road model generation method shown in an embodiment of the present application.

[0062] See also Figure 1 , the road model generation method includes:

[0063] Step S101: Acquire a first road model generated in the cloud, where the first road model at least includes: first road geometry information and first road topology information.

[0064] In this application, the cloud-based road model is a global map built by a cloud service provider. It contains information such as roads, buildings, landmarks, traffic flow, and speed limits. This cloud-based road model map information is typically used for navigation, path planning, and global positioning. Autonomous driving systems can obtain global information from the cloud-based map, such as road topology, traffic lights, and zebra crossings.

[0065] In this embodiment, the road model includes geometric and topological information. Geometric information characterizes the geometric structure of lanes. Topological information includes both the physical attributes of the user's driving road and semantic information representing the user's driving preferences. The user's driving semantics can be determined based on the topological information. The physical attributes of the user's driving road include at least one or more of lane turning information, lane type, and lane connectivity. Lane turning information indicates lane turn information, such as straight ahead, left turn, or right turn. Lane types include at least bus lanes, non-motorized vehicle lanes, tidal lanes, and roundabouts. Lane connectivity uses predecessor and successor information to indicate which lanes can be accessed without changing lanes. Semantic information regarding user driving preferences includes lane priority, lanes the user passes through when learning a route, and lanes that are farthest to the right or of poor quality, among other low-priority lanes. This topological information can be stored in corresponding lane line categories using corresponding identifiers.

[0066] In one embodiment, the first road topology information includes at least one or more of lane turning information, preceding and succeeding information, lane type, and user driving trajectory.

[0067] In this embodiment, the lane succession relationship includes, but is not limited to, being represented by an adjacency list stored in a graph, and the lane turning information is represented by a combination of symbol types, for example:

[0068] enumLaneTurnType{

[0069] TURN_TYPE_UNKNOWN=0; (indicates that the vehicle's steering is unknown);

[0070] STRAIGHT=1; (indicates the vehicle is going straight);

[0071] TURN_LEFT=2; (indicates the vehicle turns left);

[0072] TURN_RIGHT=3; (indicates the vehicle turns right);

[0073] U_TURN=4; (indicates vehicle U-turn);

[0074] STRAIGHT_AND_TURN_LEFT=5; (indicates that the vehicle turns left after going straight);

[0075] STRAIGHT_AND_TURN_RIGHT=6; (indicates that the vehicle turns right after going straight);

[0076] STRAIGHT_AND_UTURN=7; (indicates that the vehicle turns around after going straight);

[0077] STRAIGHT_AND_TURN_LEFT_AND_TURN_RIGHT=8; (indicates that the vehicle goes straight, then turns left and then right);

[0078] STRAIGHT_AND_TURN_LEFT_AND_UTURN=9; (indicates that the vehicle goes straight, turns left, and then makes a U-turn);

[0079] STRAIGHT_AND_TURN_RIGHT_AND_UTURN=10; (indicates that the vehicle goes straight, turns right, and then makes a U-turn);

[0080] TURN_LEFT_AND_TURN_RIGHT=11; (indicates that the vehicle turns left and then right);

[0081] TURN_LEFT_AND_UTURN=12; (indicates that the vehicle turns left and then makes a U-turn);

[0082] TURN_RIGHT_AND_UTURN=13; (indicates that the vehicle turns right and then turns left).

[0083] This embodiment assigns values ​​to different lane turn combinations to obtain lane turn indicators, thereby facilitating the rapid reading of lane turn information from the topological information based on the turn indicators. It should be noted that in this application, the geometric information and topological information can be combined into a unified data structure or stored in different sets, and the corresponding road information can be found by querying the corresponding set.

[0084] Figure 2 This is a schematic diagram of a road model generated in the cloud in an embodiment of the present application. Figure 2 Including lanes 1-7, connecting lines 1-4, center lines 1-7, and landmark arrows indicating turning information. Among them, each lane corresponds to a center line including lane information, lane n corresponds to center line n, for example, lane 1 corresponds to center line 1, and each center line stores the corresponding geometric feature information in the form of a chain. Connecting lines are used to represent the corresponding predecessor and successor information, and the direction of the connecting lines is used to represent the connection trajectory. For example, connecting line 1 represents the predecessor and successor information from lane 5 to lane 1, connecting line 2 represents the predecessor and successor information from lane 5 to lane 2, connecting line 3 represents the predecessor and successor information from lane 6 to lane 3, and connecting line 4 represents the predecessor and successor information from lane 7 to lane 3. It should be noted that the relevant attribute identification of the lane line is not in Figure 2 Indicated in. Figure 3 is the first road geometry information of the first road model in the embodiment of the present application, Figure 3 A simple lane map stored in the cloud. Figure 3 Lanes 1 to 7 are shown. Figure 3 The left and right boundaries of the lane line and the lane line cross-section diagram are saved.

[0085] Step S102, obtaining vehicle-sensing map information detected by the vehicle side.

[0086] Vehicle perception maps are local maps constructed using the vehicle's sensors (such as lidar, cameras, and radar). These maps contain road information surrounding the vehicle, including obstacles, lane markings, road signs, pedestrians, and other vehicles. These maps are typically used for local path planning, obstacle avoidance, and environmental awareness.

[0087] In this embodiment, the vehicle-sensing map information includes but is not limited to: lane line geometry information, lane line attributes, and lane vehicle landmarks perceived by the vehicle's sensors. Due to insufficient vehicle-side sensing capabilities, there may be problems such as missed detection of landmarks in the lane and missed detection of lane lines. It should be noted that the vehicle-sensing map information does not include the vehicle's preceding and succeeding information. Figure 4 For example, Figure 4 The vehicle sensing map information in the embodiment of the present application, ( Figure 4 The road attributes perceived by the ego vehicle’s sensors are not shown). Figure 4 The vehicle-sensing road information includes the road geometry information generated by the vehicle perception. Figure 4 and Figure 3 Indicates the same road, Figure 3 compared to, Figure 4 The lane markings of lanes 2 and 3 are missing. Figure 2 compared to, Figure 4The vehicle sensing map information in lane 1 still lacks some landmarks for lane turning information.

[0088] Step S103 : Matching the vehicle-sensing map information with the first road geometry information and the first road topology information of the first road model to obtain target road geometry information and target road topology information.

[0089] The autonomous driving system uses target road geometry and topology information to perceive the vehicle's surrounding environment in real time. This application fuses the cloud-based first road model with vehicle-sensing map information to generate a target road model, facilitating the accurate acquisition of global and local information for autonomous driving. For example, global path planning can use cloud-based maps, while local obstacle avoidance relies on vehicle-based map information.

[0090] In one embodiment, matching the vehicle-sensing map information with the first road geometry information and the first road topology information of the first road model to obtain target road geometry information and target road topology information includes: matching the vehicle-sensing map information with the first road geometry information to obtain target road geometry information; and constructing target road topology information based on the first road topology information, the vehicle-sensing map information, and the target road geometry information.

[0091] Due to factors such as vehicle-side positioning deviation and the cumulative perception error generated by vehicle-side sensors, there is a large deviation between the road model sent from the cloud and the road model constructed based on lane perception. The geometric features of the vehicle-side model need to be matched first.

[0092] In one embodiment, vehicle-sensing map information generated by a vehicle is received, and first road geometry information is matched according to the vehicle-sensing map information to obtain target road geometry information, including: obtaining second road geometry information according to the vehicle-sensing map information; geometrically matching the first road geometry information and the second road geometry information, and generating target road geometry information according to the matching result.

[0093] When the geometric structure of the cloud-based road structure is applied to the vehicle-based map, vehicle-based road information can generate discrepancies in road geometry due to vehicle positioning errors and sensor perception errors. Geometric matching and calibration based on the vehicle's real-time perception information and the primary road geometry can complement other undetected lane geometric attributes, such as the left and right boundaries of missed lanes.

[0094] In one embodiment, the first road geometric information and the second road geometric information are geometrically matched, including: determining a reliable range of self-vehicle perception, and determining target road geometric information based on the second road geometric information within the reliable range of self-vehicle perception; if the range is exceeded, correcting the second road geometric information based on the first road geometric information to determine the target road geometric information.

[0095] In this embodiment, the sensing range of the ego vehicle may be a real vehicle sensing frame. In the real vehicle sensing frame near the ego vehicle, the second road geometry information sensed by the real vehicle is more reliable, that is, at the near-vehicle end, the lane line information detected by the ego vehicle is consistent with the real road. In the far-end range of the ego vehicle beyond the real vehicle sensing frame, the road geometry information in the cloud is more trusted. Since the perception capability of the ego vehicle decreases with distance, this embodiment trusts the second road geometry information within the reliable range of ego vehicle perception, thereby ensuring greater safety and reliability near the ego vehicle. Beyond the reliable range of ego vehicle perception, the first road geometry information is used to correct the second road geometry information, thereby relying on the beyond-visual-range information in the cloud to make up for the insufficiency of vehicle-side perception, thereby achieving a strong combination of map models and ego vehicle applications. Optionally, the ego vehicle perception range is in the range of 0.5 vehicle bodies to 1.5 vehicle bodies.

[0096] In one embodiment, when the vehicle exceeds the reliable perception range, the second road geometric information is corrected based on the first road geometric information to determine the target road geometric information, including: obtaining the first road centerline of the first road geometric information and the second road centerline of the second road geometric information according to a preset geometric information conversion format; and weighted fitting the first road centerline and the second road centerline that exceed the reliable perception range of the vehicle to determine the target road geometric information.

[0097] Lane geometry information is a set of points in three-dimensional space. The default geometric information conversion format converts lane point information into an ordered series of point chains. This conversion involves converting each lane pixel into a cloud-to-vehicle coordinate system to obtain the road's three-dimensional coordinates, removing height information from these three-dimensional road coordinates to generate two-dimensional data points represented by (x, y), and then arranging and connecting these two-dimensional data points along the lane line to generate a point chain.

[0098] Since the lane centerline geometric point chain should be located at the center of the lane line, the geometric deviation of the entire lane can be corrected according to the center line of the lane line. For the point chain curve 1 representing the center line of the first road and the point chain curve 2 representing the center line of the second road, a weighted transition method is adopted between the point chain curve 1 and the point chain curve 2, specifically including: using linear interpolation to represent the weighted transition between the point chain curve 1 and the point chain curve 2, setting the weighted average formula (1), which includes:

[0099] y=w(x)y1+(1-w(x))y2 formula (1);

[0100] Here, y1 and y2 are the values ​​of point chain curve 1 and point chain curve 2 at this location, respectively. At the starting point, w(x) is 0, indicating 100% confidence in point chain curve 2. As the curve moves away from the reliable range of the vehicle, w(x) approaches 1, indicating increasing confidence in point chain curve 1.

[0101] The specific weight function w(x) can be either a linear weight function as shown in formula (2) or an exponential weight function as shown in formula (3).

[0102]

[0103] Where L is the total length of the curve.

[0104] w(x)=e -kx Formula (3);

[0105] Among them, k is a positive number used to control the decay speed of the weight and the smoothness of the transition form.

[0106] In this embodiment, when the vehicle's perception reliability range is exceeded, point chain curves 1 are selected one by one from the first set of road centerlines, and point chain curves 2 corresponding to the lanes of point chain curve 1 are selected one by one from the second set of road centerlines. Point chain curves 1 and 2 are then gradually fitted together using a preset weighting algorithm. Because dotted dot chain lines contain road feature information, fitting dot chain curves allows for rapid correction of the vehicle-side map's geometric features.

[0107] Because the vehicle-side can only perceive partial road topology information—for example, it can detect road turning information and lane line attributes—but cannot directly perceive preceding and succeeding information, the vehicle-side must reconstruct the target road topology information based on the corrected target road geometry information. In step S103, after correcting the vehicle-side road information to obtain the target road geometry information, the vehicle-side first road topology information is received from the cloud and reconstructed based on the vehicle-sensed map information.

[0108] In one embodiment, target road topology information is constructed based on first road topology information, vehicle-sensing map information, and target road geometry information, including: verifying the first road topology information, correcting the first road topology information that has not passed based on the vehicle-sensing map information, and obtaining second road topology information; comparing the vehicle-sensing map information and the first road topology information to determine newly added road topology information; and generating target road topology information based on the newly added road topology information and the second road topology information.

[0109] In this embodiment, the first topology information in the cloud is verified before use, and unreasonable first road topology information is corrected using vehicle-sensing map information collected by the vehicle.

[0110] In one embodiment, verifying the first road topology information includes: judging the rationality of the connection between the predecessor and successor information of the first road topology information according to a preset judgment rule; and verifying the accuracy of the lane turning information, predecessor and successor information, and lane type of the first road topology information according to the vehicle sensing map information.

[0111] Because lane turning information and lane type are real lane attributes in the physical world and are closely tied to landmark arrows, this embodiment uses vehicle-sensing map information to correct the lane turning information and lane type of the first road topology information within the reliable range of the vehicle's perception. Beyond the reliable range of the vehicle's perception, if the first road topology information is confirmed to be reliable, the target road topology information is constructed based on the first road topology information. It should be noted that if vehicle-sensing map information is missing, the missing parts of the vehicle-sensing map information are filled in using the first road topology information.

[0112] Previous and subsequent information represent the connectivity relationship and are used to indicate the connectivity between the current lane and the next lane. In autonomous driving, this information represents lanes with right-of-way that do not require lane changes. This information can be used to construct the road network topology, ensuring that at least one traversable path is provided to the vehicle during autonomous driving. Because the vehicle-sensing map does not include previous and subsequent information, the previous and subsequent information of the first road topology information is used to construct the target road topology model.

[0113] When using the preceding and succeeding information from the cloud, verification of the preceding and succeeding information is required. Preset judgment rules include, but are not limited to, evaluating the smoothness of the road after the preceding and succeeding information are connected, and / or evaluating whether the road connected by the preceding and succeeding information complies with road regulations, and / or whether the preceding and succeeding information crosses the road boundary.

[0114] In this embodiment, if the preceding and succeeding information do not conform to the rules, the preceding and succeeding information between the preceding and succeeding roads is regenerated based on the center lines of the preceding and succeeding roads.

[0115] In one embodiment, comparing the vehicle-sensing map information and the first road topology information to determine the newly added road topology information includes: obtaining the newly added user driving trajectory and the newly added user driving habits based on the vehicle-sensing map information; and generating the newly added road topology information based on the newly added user driving trajectory and the newly added user driving habits.

[0116] This embodiment collects new user driving routes and behaviors in real time based on vehicle-sensing map information. After determining the newly added user's driving trajectory and driving habits, the relevant data is automatically transmitted back in its entirety. After screening and verification, the newly added user's driving trajectory and driving habits data are transmitted back to the cloud server. This transmitted data contributes to the construction of the target vehicle-side road model and also modifies the relevant road data in the cloud. The next time the user drives the same route, the first road model containing the newly added data is automatically pulled from the cloud, and the relevant designated driving service begins.

[0117] This embodiment converts the newly added user's driving trajectory and newly added user's driving habits into newly added road topology information, so that the target road model completes the learning of the user's driving route and driving behavior. After all the learning is completed, the newly added road topology information is automatically transmitted back to the full segment data. After screening and inspection, it is transmitted back to the cloud server to execute subsequent cloud map construction. The next time the user drives and executes the same route, he can activate the function, and the vehicle will automatically pull the already constructed map from the cloud and start driving for him.

[0118] In one embodiment, the newly added road topology information is generated based on the newly added user's driving trajectory and the newly added user's driving habits, including: determining the priority of the driving lane and / or the priority of the lane changing lane based on the newly added user's driving trajectory and the newly added user's driving habits; generating an identifier of the priority of the driving lane and / or the priority of the lane changing lane to determine the newly added road topology information.

[0119] This embodiment extracts user driving habit features from a user's driving trajectory and converts these into newly added topological information. For example, based on the user's driving trajectory, it can determine which lane the user prefers and where they prefer to change lanes. The user's preferred lane and lane location are stored in the newly added road topological information as priority identifiers, making it easier for the autonomous driving system to subsequently tailor navigation behavior to the user's habits. In this embodiment, acquiring user driving habit features specifically includes: driving from the departure point to the destination according to the user's set departure and destination points, and recording the vehicle's trajectory. Using environmental perception technology to record road boundary information during the learning process, the user's preferred lane and driving habits (e.g., driving left or right within a lane) are determined based on this road boundary information. The user's preferred lane and lane marking are then labeled accordingly and stored in the newly added topological information. It should be noted that determining the user's preferred lane and driving habits can be accomplished by learning user trajectories using relevant AI models, or by statistically analyzing user trajectories to determine newly added user driving preferences.

[0120] In one embodiment, after the newly added road topology information is determined, the method further includes: modifying the first road topology information of the first road model according to the newly added road topology information.

[0121] This embodiment modifies the first road topology information so that when the user pulls the first road model generated in the cloud next time, the first road model retains the driving semantics of the newly added user.

[0122] Step S104: Generate a vehicle-side target road model based on the target road geometry information and the target road topology information.

[0123] The target road model of this application relies on the beyond-visual-range advantage of the cloud. First, a high-quality first road model is generated in the cloud. The first road model includes basic road geometry and topology. Then, the necessary road geometry and topology information is sent to the vehicle side. When the vehicle side consumes the prior information, it performs real-time geometric matching of the geometry with the vehicle-sensing map information perceived by the real vehicle, thereby constructing the target road geometry features. The target road geometry features have a more stable road structure, reducing the instability of the road structure caused by the large volatility of the vehicle-side perception information. At the same time, the first road topology information with prior information in the cloud is assigned to the vehicle-side road structure to solve the road information loss caused by insufficient vehicle-side perception capabilities. Since the prior road structure information in the cloud is consumed in the construction of the target road topology information, memory and time resources can be saved during the construction process.

[0124] Figure 5 Schematic diagram of the target road model of an embodiment of the present application. Figure 2 for Figure 5 The corresponding first road model, Figure 4 for Figure 5 Corresponding car map information. Combined Figures 2 to 5 , the target road model constructed in this embodiment is described. Figure 2 The first road geometry information in Figure 3 , get Figure 4 The second road geometry information is matched with the first road geometry information, and the second road geometry information is corrected according to the matching result. Figure 3 ,get Figure 5 The road structure (that is, the target road geometry information). Figure 5 The target road topology information is reconstructed based on the road structure of the vehicle, specifically including: obtaining first road topology information, verifying the first road topology information according to a preset rule, obtaining second road topology information, and filling in the missing road turning information in the vehicle sensing map information according to the second road topology information ( Figure 5The left turn arrow in lane 1 is the road turning information filled in by the second road information). At the same time, the driving trajectory and driving habits of the newly added users detected by the vehicle are collected, and the newly added lane topology information is generated based on the driving trajectory and driving habits of the newly added users, and the newly added lane topology information is added to the Figure 5 , the target road model is obtained. It should be noted that the newly added user's driving trajectory and driving habits are not only used to construct the target road model on the vehicle; they are also used to modify or overwrite the user's historical trajectory and driving habits stored in the cloud. This ensures that when the user pulls the first road model from the cloud the next time they drive, the first road topology information of the first road model will include the newly added user's driving trajectory and driving habits, thus ensuring the accuracy of the driving semantics of the next vehicle-side target map.

[0125] The technical solution provided by this application may have the following beneficial effects:

[0126] First, this application acquires a cloud-generated first road model and vehicle-based vehicle-sensing map information. Based on the a priori information of the first road model, the vehicle-based map information is integrated to generate a target road model on the vehicle. Compared to traditional methods of generating vehicle-based road models, this application incorporates prior information from the cloud into the vehicle-based road structure, saving computing power on the vehicle side to build the road model and improving the efficiency of generating the target road model.

[0127] Secondly, this application retains the cloud's advantage in building road models beyond visual range by consuming the first road geometry information and the first road topology information in the cloud; by matching the vehicle-sensing map information, the first road geometry information, and the first road topology information, it can reduce the instability of the road structure caused by the large volatility of the vehicle-side perception information, thereby building a more stable target road model. At the same time, the vehicle-sensing map information can collect the user's newly added driving trajectory in real time, and by integrating the user's newly added driving trajectory into the target road model, the target road model can update the user's driving characteristics in real time.

[0128] like Figure 6 As shown, a road model generation method and device includes:

[0129] A first acquisition module 601 is configured to acquire a first road model generated in the cloud, wherein the first road model includes at least first road geometry information and first road topology information;

[0130] The second acquisition module 602 is used to obtain vehicle-sensing map information detected by the vehicle side;

[0131] A processing module 603 is configured to match the vehicle-sensing map information with first road geometry information and first road topology information of the first road model to obtain target road geometry information and target road topology information;

[0132] The determination module 604 generates a vehicle-side target road model based on the target road geometry information and the target road topology information.

[0133] In one embodiment, the processing module also includes a first processing module and a second processing module. The first processing module is used to match the vehicle-sensing map information with the first road geometry information to obtain the target road geometry information; the second processing module is used to construct the target road topology information based on the first road topology information, the vehicle-sensing map information and the target road geometry information.

[0134] In one embodiment, the first road topology information includes at least one or more of lane turning information, preceding and succeeding information, lane type, and user driving trajectory.

[0135] In one embodiment, the vehicle-sensing map information is matched with the first road geometry information to obtain the target road geometry information, and the target road geometry information is obtained, including: obtaining the second road geometry information based on the vehicle-sensing map information; geometrically matching the first road geometry information and the second road geometry information, and generating the target road geometry information based on the matching result.

[0136] In one embodiment, the first road geometric information and the second road geometric information are geometrically matched, including: determining a reliable range of self-vehicle perception, and determining target road geometric information based on the second road geometric information within the reliable range of self-vehicle perception; if the range is exceeded, correcting the second road geometric information based on the first road geometric information to determine the target road geometric information.

[0137] In one embodiment, when the vehicle exceeds the reliable perception range, the second road geometric information is corrected based on the first road geometric information to determine the target road geometric information, including: obtaining the first road centerline of the first road geometric information and the second road centerline of the second road geometric information according to a preset geometric information conversion format; and weighted fitting the first road centerline and the second road centerline that exceed the reliable perception range of the vehicle to determine the target road geometric information.

[0138] In one embodiment, target road topology information is constructed based on first road topology information, vehicle-sensing map information, and target road geometry information, including: verifying the first road topology information, correcting the first road topology information that has not passed the verification based on the vehicle-sensing map information, and obtaining second road topology information; comparing the vehicle-sensing map information and the first road topology information to determine newly added road topology information; and generating target road topology information based on the newly added road topology information and the second road topology information.

[0139] In one embodiment, verifying the first road topology information includes: judging the rationality of the connection between the predecessor and successor information of the first road topology information according to a preset judgment rule; and / or verifying the accuracy of the lane turning information, predecessor and successor information and lane type of the first road topology information according to the vehicle sensing map information.

[0140] In one embodiment, comparing the vehicle-sensing map information and the first road topology information to determine the newly added road topology information includes: obtaining the newly added user's driving trajectory and the newly added user's driving habits based on the vehicle-sensing map information; and generating the newly added road topology information based on the newly added user's driving trajectory and the newly added user's driving habits.

[0141] Figure 7 It is a schematic structural diagram of a car shown in an embodiment of the present application.

[0142] See also Figure 7 , the car 700 includes a memory 710 and a processor 720 .

[0143] The processor 720 may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor.

[0144] The memory 710 may include various types of storage units, such as system memory, read-only memory (ROM), and permanent storage. ROM may store static data or instructions required by the processor 720 or other modules of the computer. The permanent storage may be a readable and writable storage device. The permanent storage may be a non-volatile storage device that retains stored instructions and data even when the computer is powered off. In some embodiments, the permanent storage device uses a large-capacity storage device (e.g., a magnetic or optical disk, flash memory) as the permanent storage device. In other embodiments, the permanent storage device may be a removable storage device (e.g., a floppy disk, optical drive). The system memory may be a readable and writable storage device or a volatile readable and writable storage device, such as dynamic random access memory. The system memory may store some or all instructions and data required by the processor during operation. In addition, the memory 710 may include any combination of computer-readable storage media, including various types of semiconductor memory chips (e.g., DRAM, SRAM, SDRAM, flash memory, programmable read-only memory), and magnetic disks and / or optical disks may also be used. In some embodiments, the memory 710 may include a readable and / or writable removable storage device, such as a compact disc (CD), a read-only digital versatile disc (e.g., DVD-ROM, dual-layer DVD-ROM), a read-only Blu-ray disc, an ultra-density optical disc, a flash memory card (e.g., SD card, mini SD card, Micro-SD card, etc.), a magnetic floppy disk, etc. Computer-readable storage media do not include carrier waves and transient electronic signals transmitted wirelessly or wired.

[0145] The memory 710 stores executable codes. When the executable codes are processed by the processor 720 , the processor 720 may execute part or all of the above-mentioned methods.

[0146] In addition, the method according to the present application may also be implemented as a computer program or a computer program product, which includes computer program code instructions for executing some or all of the steps in the above method of the present application.

[0147] Alternatively, the present application can also be implemented as a computer-readable storage medium (or non-transitory machine-readable storage medium or machine-readable storage medium) on which executable code (or computer program or computer instruction code) is stored. When the executable code (or computer program or computer instruction code) is executed by the processor of a car (or server, etc.), the processor executes part or all of the steps of the above-mentioned method according to the present application.

[0148] The embodiments of the present application have been described above. The above description is exemplary, not exhaustive, and is not limited to the disclosed embodiments. Many modifications and variations will be apparent to those skilled in the art without departing from the scope and spirit of the described embodiments. The terminology used herein is selected to best explain the principles of the embodiments, their practical applications, or improvements to the technology in the market, or to enable other persons skilled in the art to understand the embodiments disclosed herein.

Claims

1. A road model generation method, characterized in that: include: Obtaining a first road model generated in the cloud, where the first road model includes at least: first road geometry information and first road topology information; Obtain vehicle-sensing map information detected by the vehicle side; Matching the vehicle-sensing map information with first road geometry information and first road topology information of the first road model to obtain target road geometry information and target road topology information; wherein, second road geometry information is obtained based on the vehicle-sensing map information; determining a reliable range of self-vehicle perception, and determining the target road geometry information based on the second road geometry information within the reliable range of self-vehicle perception; if the reliable range of self-vehicle perception is exceeded, obtaining a first road centerline of the first road geometry information and a second road centerline of the second road geometry information according to a preset geometry information conversion format; performing weighted fitting on the first road centerline and the second road centerline that exceed the reliable range of self-vehicle perception to determine the target road geometry information; constructing target road topology information according to the first road topology information, the vehicle sensing map information, and the target road geometry information; A vehicle-side target road model is generated based on the target road geometry information and the target road topology information.

2. The method according to claim 1, characterized in that The first road topology information includes at least one or more of lane turning information, preceding and succeeding information, lane type, and user driving trajectory.

3. The method according to claim 1, characterized in that The point chain curve 1 is used to represent the center line of the first road, the point chain curve 2 is used to represent the center line of the second road, and linear interpolation is used to represent the weighted transition between the point chain curve 1 and the point chain curve 2; the weighted average formula (1) is set, and the formula (1) includes: y=w(x)y1+(1-w(x))y2 (1); Among them, y1 and y2 are the values ​​of point chain curve 1 and point chain curve 2 corresponding to the current coordinate position x; The weight function w(x) is selected from the linear weight function shown in formula (2) or the exponential weight function shown in formula (3); Where L is the total length of the curve; w(x)=e -kx (3); Among them, k is a positive number used to control the decay speed of the weight and the smoothness of the transition form.

4. The method according to claim 1, wherein The constructing target road topology information according to the first road topology information, the vehicle sensing map information, and the target road geometry information includes: verifying the first road topology information, and correcting the first road topology information that fails to pass the verification according to the vehicle-sensing map information to obtain second road topology information; Comparing the vehicle-sensing map information with the first road topology information to determine newly added road topology information; The target road topology information is generated according to the newly added road topology information and the second road topology information.

5. The method according to claim 4, characterized in that The verifying the first road topology information includes: Determining the rationality of the connection between the preceding and succeeding information of the first road topology information according to a preset judgment rule; and / or; Verify the accuracy of lane turning information, preceding and following information, and lane type of the first road topology information based on the vehicle sensing map information.

6. The method according to claim 4, characterized in that The comparing the vehicle-sensing map information and the first road topology information to determine the newly added road topology information includes: Obtaining the driving trajectory and driving habits of the newly added user based on the vehicle sensing map information; The newly added road topology information is generated according to the newly added user's driving trajectory and the newly added user's driving habits.

7. The method according to claim 1, characterized in that The geometric information of lane lines is a set of points in three-dimensional space; The preset geometric information conversion format converts lane line point set information into a series of ordered point chain sets, wherein converting the three-dimensional point set information into the point chain set includes: The pixel points on each lane line are converted through the cloud-vehicle coordinate system to obtain the three-dimensional coordinate points of the road. The height information of the three-dimensional coordinate points of the road is removed to generate two-dimensional data points represented by (x, y). The two-dimensional data points are arranged and connected according to the direction of the lane line to generate a point chain set.

8. The method according to claim 7, characterized in that The point chain curve 1 is used to represent the center line of the first road, and the point chain curve 2 is used to represent the center line of the second road; When the vehicle exceeds the reliable perception range, point chain curve 1 is selected one by one from the first road centerline set, and point chain curve 2 corresponding to the lane of point chain curve 1 is selected one by one from the second road centerline set. According to a preset weighted algorithm, the point chain curve 1 and the point chain curve 2 are gradually fitted, the second road geometric information of the vehicle-side map is corrected, and the target road geometric information is determined.

9. The method according to claim 6, characterized in that The obtaining of the newly added user's driving trajectory and the newly added user's driving habits includes: According to the departure and destination set by the user, the vehicle travels from the departure point to the destination and records the vehicle's travel trajectory.

10. A road model generating device, characterized in that: include: A first acquisition module is configured to acquire a first road model generated in the cloud, wherein the first road model includes at least first road geometry information and first road topology information; The second acquisition module is used to obtain vehicle-sensing map information detected by the vehicle side; a processing module for matching the vehicle-sensing map information with first road geometry information and first road topology information of the first road model to obtain target road geometry information and target road topology information; wherein, second road geometry information is obtained based on the vehicle-sensing map information; a reliable range of self-vehicle perception is determined, and within the reliable range of self-vehicle perception, the target road geometry information is determined based on the second road geometry information; if the reliable range of self-vehicle perception is exceeded, a first road centerline of the first road geometry information and a second road centerline of the second road geometry information are obtained according to a preset geometry information conversion format; weighted fitting of the first road centerline and the second road centerline that exceed the reliable range of self-vehicle perception is performed to determine the target road geometry information; and target road topology information is constructed based on the first road topology information, the vehicle-sensing map information, and the target road geometry information; The determination module generates a vehicle-side target road model based on the target road geometry information and the target road topology information.

11. An automobile, characterized in that: include: processor; as well as A memory having executable codes stored thereon, which, when executed by the processor, causes the processor to perform the method according to any one of claims 1 to 9.

12. A computer-readable storage medium having executable code stored thereon, wherein when the executable code is executed by a processor of an electronic device, the processor is caused to execute the method according to any one of claims 1 to 9.

Images