Method of generating high-resolution environment map for vehicle having autonomous travel function
By using a knowledge graph to supplement sensor data with domain knowledge, the method addresses HD map challenges, enabling accurate and real-time creation and update of high-resolution maps for autonomous vehicles, improving navigation in dynamic environments.
Patent Information
- Application Number
- JP2025030105
- Authority / Receiving Office
- JP · JP
- Patent Type
- Applications
- Current Assignee / Owner
- Priority Date
- 2024-02-28
- Filing Date
- 2025-02-27
- Publication Date
- 2025-09-09
AI Technical Summary
Existing high-definition (HD) maps for autonomous vehicles face challenges due to high production costs, update difficulties, limited availability, and accuracy issues in dynamic environments, necessitating improved methods for creating and updating these maps.
A method using a knowledge graph to supplement environmental image data from vehicle sensors with domain knowledge, enabling accurate recognition and classification of map elements, and incorporating spatial relationships to create high-resolution environment maps.
Enables accurate, real-time HD map creation and update, particularly in areas with limited access, providing a more detailed and up-to-date map source for autonomous vehicles, enhancing their navigation capabilities.
Smart Images
Figure 2025131554000001_ABST
Abstract
Description
[Technical Field]
[0001] The present invention relates to a method and apparatus for creating a high-resolution environment map for a vehicle with autonomous driving capabilities. [Background technology]
[0002] Conventional technology Autonomous driving systems (AD) represent the pinnacle of technological innovation in the automotive sector, promising to revolutionize mobility and dramatically improve safety and efficiency on public roads. A core component essential for these systems to function optimally is high-definition (HD) maps. These maps provide autonomous vehicles with detailed information about their driving environment, enabling precise localization and supporting decision-making in complex traffic situations. However, despite their importance, HD maps present significant hurdles due to their high production costs, the challenges of keeping them updated, and their limited availability in certain regions. Furthermore, the accuracy of these maps can be limited in dynamic environments where road conditions change rapidly.
[0003] In this context, creating and continuously updating HD maps using the perception systems of autonomous vehicles opens up a promising approach: if vehicles could capture their surroundings in real time, collect data, and use this information to refine their maps, many of the existing challenges could be overcome.
[0004] The academic paper "VectorMapNet: End-to-End Vectorized HD Map Learning," by Y. Liu, Y. Yuan, Yue. Wang, Y. Wang, H. Zhao et al. (submitted to ICLR2023, 2022) [https: / / arxiv.org / abs / 2206.08920], addresses the problem of creating high-resolution (HD) maps from onboard sensors for autonomous driving. This method uses observational data from onboard sensors, such as cameras, lidar, and radar, to predict polylines representing various map elements, such as lanes, crosswalks, or lane divisions. This method essentially consists of three steps: In feature extraction, image features are extracted using a ResNetCNN, followed by mapping the image to a bird's-eye view (BEV) using inverse perceptual mapping (IPM). Furthermore, the lidar observations are processed into PointPillars using dynamic voxelization. Similarly, the image features and lidar features are also interconnected and subsequently processed using a two-layer CNN. Furthermore, a map element detector is also disclosed, in which a transformer ensemble predictive detector (DETR) recognizes element keypoints that subsequently form polylines of map elements. A deformable attention module is also used, in which each element query has a unique localization. The prediction head includes two MLPs that decode element queries into element keypoints and their class labels. Furthermore, in the third step, a polyline generator is used, which generates detailed geometric shapes from multiple map elements by modeling the distribution of map element vertices and BEV features.
[0005] Further prior art is known from the scientific publication "HDMapGen: Hierarchical Graph Generative Model of High Definition Maps, IEEE CVPR, 2021" by L. Moi et al. [Prior art documents] [Non-patent literature]
[0006] [Non-Patent Document 1] Y. Liu, Y. Yuan, Yue. Wang, Y. Wang, H. Zhao et al., academic paper "VectorMapNet: End-to-End Vectorized HD Map Learning," https: / / arxiv.org / abs / 2206.08920 (submitted to ICLR2023, 2022) [Non-patent document 2] Scientific publication by L. Moi et al., "HDMapGen: Hierarchical Graph Generative Model of High Definition Maps, IEEE CVPR, 2021" Summary of the Invention [Problem to be solved by the invention]
[0007] The problem on which the present invention is based is to provide an improved method and / or device for creating a high-resolution environment map for a vehicle with autonomous driving capabilities. [Means for solving the problem]
[0008] The problem is solved by a method according to the characterizing part of claim 1. The problem is solved by a method according to the characterizing part of claim 2. The problem is solved by an apparatus according to the characterizing part of claim 10. The problem is solved by an apparatus according to the characterizing part of claim 11.
[0009] Disclosure of the Invention According to a first aspect, there is provided a method for creating a high-resolution environment map (also referred to as an HD map) for an autonomous vehicle (also referred to as an AD vehicle), the method comprising: - providing environmental image data for a perception system of the vehicle, the perception system having an environmental sensor for capturing the vehicle environment while the vehicle is moving; providing domain knowledge of the vehicle environment in the form of a trained knowledge graph; - creating a high-resolution environment map of the vehicle environment by supplementing the environmental image data with the provided domain knowledge of the trained knowledge graph; Includes:
[0010] According to a second aspect, a method for generating a high-resolution environment map for a vehicle having autonomous driving capabilities is provided, the method comprising: - optionally providing environmental image data of a perception system of the vehicle, the perception system having an environmental sensor for capturing the vehicle environment while the vehicle is moving; providing a high-resolution environment map of a vehicle environment; - providing domain knowledge of the vehicle environment in the form of a trained knowledge graph; - validating and / or supplementing high resolution environment maps and / or optionally environmental image data of the vehicle environment with the provided domain knowledge of the trained knowledge graph; Includes:
[0011] It is understood that the steps according to the invention and further optional steps do not necessarily have to be performed in the order shown and can be performed in a different order. Furthermore, further intermediate steps may be provided. Furthermore, the individual steps may comprise one or more sub-steps without thereby departing from the scope of the method according to the invention according to the first or second aspect.
[0012] According to a third aspect, an apparatus is presented for generating a high-resolution environment map for a vehicle having autonomous driving capabilities, the apparatus comprising: The apparatus includes an evaluation and calculation unit; The evaluation and calculation device comprises the following steps: - providing environmental image data for a perception system of the vehicle, the perception system having an environmental sensor for capturing the vehicle environment while the vehicle is moving; providing domain knowledge of the vehicle environment in the form of a trained knowledge graph; - creating a high-resolution environment map of the vehicle environment by supplementing the environmental image data with the provided domain knowledge of the trained knowledge graph; The system is configured to:
[0013] According to a fourth aspect, there is provided an apparatus for generating a high resolution environment map for a vehicle having autonomous driving capabilities, the apparatus comprising: The apparatus includes an evaluation and calculation unit; The evaluation and calculation device comprises the following steps: - optionally providing environmental image data of a perception system of the vehicle, the perception system having an environmental sensor for capturing the vehicle environment while the vehicle is moving; providing a high-resolution environment map of a vehicle environment; - providing domain knowledge of the vehicle environment in the form of a trained knowledge graph; - validating and / or supplementing high resolution environment maps and / or optionally environmental image data of the vehicle environment with the provided domain knowledge of the trained knowledge graph; The system is configured to:
[0014] The embodiments made for the method apply correspondingly to the device, where it is understood that linguistic modifications of features formulated according to the method can be reformulated for the device according to usual linguistic practices, without the need for this type of formulation to be explicitly mentioned here.
[0015] The present invention represents map elements by a knowledge graph (also known as a Knowledge-Graph (KG)), which can represent many map elements. Furthermore, the knowledge graph preferably also includes relationships between entities. The knowledge graph can be based on a standardized ontology, for example, the ASAM OpenX ontology. Therefore, the knowledge graph can represent highly detailed map elements that are important for autonomous driving or autonomous driving functionality, such as various types of lane markings, road signs, traffic lights, pedestrian walkways, parking areas, debris, traffic cones, road construction elements, and / or stop lines.
[0016] The knowledge graph preferably learns the distribution of map elements from a training dataset of labeled map elements, which can be derived or provided from a dataset for autonomous driving. Incorporating domain knowledge in the knowledge graph reduces the construction of unrealistic road elements, such as lane straightness, lane dividers, road boundaries, maximum road curvature, minimum and / or maximum lane widths, etc.
[0017] The knowledge graph is constructed from a training dataset to learn the spatial relationships between map elements and ensure that only spatially valid elements are constructed, e.g., a crosswalk begins and ends at the opposite road boundary, a lane divider exists between two adjacent lanes, and / or a road boundary exists in the outermost lane, etc.
[0018] Here, it is advantageous for the creation of a high-resolution environment map if the vehicle's perception system recognizes the vehicle environment, for example, by a lidar sensor, and segments and classifies the vehicle environment into individual map elements. Segmentation can be performed, in particular, from captured images pixel by pixel. Classification can be performed, for example, by classes such as road, roadside, building, pedestrian path, etc. By incorporating domain knowledge, on the one hand, validation of the classified map elements can be performed, in particular by assigning element-specific probabilities. This allows for the knowledge graph to also incorporate captured context knowledge, for example, regarding whether the vehicle is in an urban area, a rural road, or a highway. For example, if the vehicle is on a highway and a pedestrian crossing is segmented and classified as a map element by the environmental sensor or the used segmentation and classification algorithm, this result can be taken into account by assigning a low probability to the actual presence of a pedestrian crossing in the further creation of the high-resolution environment map by incorporating domain knowledge from the knowledge graph for the further creation of the high-resolution environment map. This results in a more accurate high-resolution environment map created in this way. The same can also be done, for example, by dimensional inspection of the recognized road width or lane width.
[0019] At least the method according to the first aspect or a corresponding device (which may be part of a system) can be used to automatically create an HD map inside a vehicle with autonomous driving capabilities (also called an AD vehicle). The AD vehicle here creates an HD map for each location it travels through. The created HD map can preferably be transmitted online to a location, e.g., a server or a cloud, where the HD maps of multiple AD vehicles are received and collected, so that a particularly complex and information-rich HD map can be created from multiple individual HD maps.
[0020] At least the method according to the first aspect or a corresponding device can also be used in AD vehicles that have no or limited access to HD maps, in particular to create HD maps instantly or in real time while driving, based on environmental image data from a recognition or perception system or a description system provided in the form of a knowledge graph, potentially making it possible to provide online HD map creation even for automated vehicles of levels 2 to 5 that previously only used SD cards.
[0021] At least the method according to the second aspect or a corresponding device can also be used to provide a more up-to-date, more accurate, and / or alternative source of information for the final HD map. This advantageously allows online HD map generation to be offered to automated vehicles, for example, Level 2 to Level 5, that already use stored HD maps. In this example, it is possible to provide a second or alternative knowledge source with potentially more accurate or more up-to-date HD map information. Furthermore, the method allows HD maps to be provided for areas not covered by stored HD maps or to adjust HD maps using supplements to existing HD maps. Thus, online HD map generation for comprehensive offline HD map creation and supplementation is also provided.
[0022] According to one embodiment, the environmental image data is validated and / or supplemented based on Standard Definition (SD) topology data of the vehicle environment.
[0023] A local topology map may comprise a map typically used in a navigation system from which road guidance and driving directions can be obtained. The map preferably exists in the form of standard definition topology data. SD map topology is preferably used to provide guide track topology, driving directions, and other information for creating and validating HD maps.
[0024] According to one embodiment, the environmental image data is collected by the following steps: - segmenting the image data of the vehicle environment captured by the environmental sensors into elements and classifying said elements into predefined classes that correspond in particular to the ontology of the knowledge graph; - transferring the segmented and classified elements of the vehicle environment to a bird's-eye view; - extracting individual elements of the vehicle environment based on the segmented image data, and / or - combining the extracted elements to create an environment map, in particular by validation using the domain knowledge provided by the knowledge graph. is pretreated by at least one of
[0025] The transition to a bird's-eye view can also be performed using image data already captured before segmentation and classification.
[0026] In this example, it is proposed to construct an HD map online in an AD vehicle. According to the first and third aspects, this construction is preferably based on environmental image data of the AD perception system, possibly based on a local topology (SD) map, and further based on domain knowledge of a trained map knowledge graph. The AD perception system is preferably used here to recognize moving objects such as vehicles and pedestrians, but also map elements such as lanes, crosswalks, etc. A preferred image segmentation is used to divide all elements into defined classes. The local topology map in this case preferably serves as a starting point for HD map construction by including the most important information about lane topology, driving direction, curves, etc. The trained map knowledge graph is used to provide information about typical specific distributions, relationship distributions, topology distributions, area-dependent distributions, and statistical distributions of map elements. This trained map knowledge graph here assists in constructing a highly accurate HD map.
[0027] The HD map building process preferably accepts individual map element candidates from a recognition or perception system and verifies their existence in a subsequent step. This is preferably done, inter alia, by connecting map elements later and hierarchically verifying their existence based on their most recent location. Temporal information can potentially be used to filter moving objects, such as vehicles and pedestrians, from the image data.
[0028] The individual map elements are preferably interconnected based on their type and their location, which are preset by the perception system, and more preferably based on information provided by the knowledge graph. Techniques such as neural networks that can process heterogeneous graphs such as knowledge graphs can preferably be used to find compact vector-based representations of traffic scenes. These can be used to calculate the similarity between the proposed HD map and the stored map knowledge graph representation.
[0029] According to one embodiment, creating a high-resolution environment map of the vehicle environment includes augmenting the environment map with domain knowledge provided by a knowledge graph.
[0030] The HD map is preferably supplemented with information that may not be visible in the image data of the environmental sensors and / or that is not provided by the perception system. For example, traffic rules, traffic lights or traffic signs usually estimate stopping areas where a vehicle should stop if it has to avoid them. Therefore, information about lane separation that is not visible can also be supplemented. In this way, the method or device can infer the valid lane separation even in the case of road works with permanent white and temporary yellow overtaking markings.
[0031] According to one embodiment, the environmental sensors include lidar and / or radar sensors and / or cameras.
[0032] The vehicle's perception system may include, for example, a video camera, a radar sensor, a lidar sensor, or other sensors. The vehicle environment may also be captured by a video sensor or a video camera alone. While the vehicle is traveling, the perception system recognizes objects in the captured driving scene and segments the entire image, preferably frame by frame, into predefined classes, particularly according to the ontology described by the knowledge graph. These elements are preferably displayed using a bird's-eye view (BEV) that corresponds to a high-resolution environment map display.
[0033] According to one embodiment, the knowledge graph is trained based on a training dataset of multiple driving scenes and / or domain knowledge.
[0034] Through this training, the knowledge graph preferably learns typical, legal, and / or reasonable representations of map elements from a training dataset. This supports the creation of more accurate high-definition (HD) maps herein. The map knowledge graph is preferably created from a training dataset including multiple driving scenes. In doing so, a map ontology is preferably created. The knowledge graph preferably shows typical distributions of map elements and their topology. The knowledge graph is preferably used as a guide for the process of online HD map creation.
[0035] According to one embodiment, the knowledge graph is trained to form spatial relationships between map elements and ensure that only elements that reasonably match each other spatially are used to create and / or supplement the high-resolution environment map.
[0036] According to one embodiment, the knowledge graph comprises domain knowledge about road types and / or lane types and / or lane dividers and / or road boundaries and / or crosswalks and / or stop areas and / or traffic signs and / or traffic lights and / or directional arrows and / or utility poles and / or barriers and / or traffic cones and / or buildings and / or plants and / or debris.
[0037] The proposed method incorporates domain knowledge stored in a trained knowledge graph to consider more map elements than previous approaches. While previous approaches are often only able to recognize, segment, classify, and therefore consider four element types in image data, the proposed method incorporates domain knowledge in a trained knowledge graph to consider multiple map elements when creating, supplementing, or validating a high-resolution environment map. Exemplary map elements include various lane types, such as vehicle lanes, bicycle lanes, pedestrian lanes, and parking lanes; various lane dividers, such as continuous lines, double continuous lines, dashed lines, dedicated dashed lines, and timed lane dividers; stop areas that are not visible but can be derived from the ontology of the knowledge graph; traffic signs, traffic lights, directional arrows attached to the road, poles, barriers, traffic cones, buildings, plants, and debris. The proposed approach can also potentially derive traffic rules from the recognized map details, such as right-of-way rules that depend on road signs and similar traffic rules.
[0038] According to the present invention, a control device is also claimed that is included in a vehicle and / or a robotic system and / or an industrial machine having an autonomous driving function and that is capable of performing any one of the methods according to the present invention.
[0039] According to the invention, a computer program is also claimed, comprising a program code for performing at least part of the method according to the invention in its embodiments, when the computer program is run on a computer. In other words, according to the invention, a computer program product is claimed, comprising instructions for causing a computer to perform the method / steps of the method according to the invention in its embodiments, when the program is run by the computer.
[0040] According to the invention, a computer-readable data carrier is also proposed, which comprises a program code of a computer program for performing at least part of the inventive method in one of its embodiments, when the computer program is run on a computer. In other words, the invention relates to a computer-readable (storage) medium which comprises instructions which, when run by a computer, cause the computer to perform the method / steps of the inventive method in one of its embodiments.
[0041] The described embodiments and developments can be combined with one another in any combination.
[0042] Further possible embodiments, developments and implementations of the invention also include combinations not expressly mentioned of the features of the invention described above or below with reference to the examples.
[0043] The accompanying drawings should facilitate a better understanding of embodiments of the present invention, and the drawings illustrate embodiments and, together with the description, serve to explain the principles and concepts of the present invention.
[0044] Other embodiments and many of the mentioned advantages will become apparent in conjunction with the drawings, in which elements shown are not necessarily drawn to scale relative to each other. [Brief explanation of the drawings]
[0045] [Figure 1] 1 is a flow chart that schematically illustrates the method according to the first aspect; [Figure 2] 3 is a flow chart that schematically illustrates the method according to the second aspect. [Figure 3] FIG. 2 is a block circuit diagram that schematically illustrates the method according to one embodiment.
[0046] In these drawing depictions, like reference numbers, unless otherwise stated, indicate identical or functionally identical elements, parts or components. DETAILED DESCRIPTION OF THE INVENTION
[0047] FIG. 1 shows a schematic flow chart of a method for creating a high-resolution environment map for a vehicle with autonomous driving capabilities according to a first embodiment.
[0048] FIG. 2 shows a schematic flow chart of a method for creating a high-resolution environment map for a vehicle with autonomous driving capabilities according to a second embodiment.
[0049] The method according to the first and second aspects can in any embodiment be performed at least in part by the device 100, and for this purpose the method may comprise several components not shown in more detail, such as one or more providing devices and / or at least one evaluation and calculation device. It is understood that the providing device may be configured together with the evaluation and calculation device or may be distinct from the evaluation and calculation device. Furthermore, the system may comprise a storage device and / or an output device and / or a display device and / or an input device.
[0050] The computer-implemented method according to the first aspect, shown in FIG. 1, according to the present invention, includes at least the following steps:
[0051] In step S1, a step of providing environmental image data for a perception system of a vehicle, where the perception system has an environmental sensor for capturing the vehicle environment while the vehicle is traveling, is performed.
[0052] In step S2, a step of providing domain knowledge of the vehicle environment in the form of a trained knowledge graph is performed.
[0053] In step S3, a step of creating a high resolution environment map of the vehicle environment is performed by supplementing the environmental image data with the provided domain knowledge of the trained knowledge graph.
[0054] The computer-implemented method according to the second aspect, shown in FIG. 2, according to the present invention, includes at least the following steps:
[0055] In optional step S10, there is provided environmental image data for a perception system of the vehicle, the perception system having environmental sensors for capturing the vehicle environment while the vehicle is moving.
[0056] In step S11, a step of providing a high resolution environment map of the vehicle environment is performed.
[0057] In step S12, a step of providing domain knowledge of the vehicle environment in the form of a trained knowledge graph is performed.
[0058] In step S13, the provided domain knowledge of the trained knowledge graph is used to validate and / or supplement the high resolution environment map and / or optionally the environment image data of the vehicle environment.
[0059] FIG. 3 shows a schematic block circuit diagram of the method according to one embodiment.
[0060] Image data 300 of the vehicle environment is captured by environmental sensors, such as lidar sensors, radar sensors, and / or cameras. The image data 300 of the vehicle environment captured by the environmental sensors is segmented into (map) elements and classified into predefined classes that correspond, in particular, to the ontology of the knowledge graph 301. These segmented and classified map elements are preferably transferred to a bird's-eye view representation 302. Extraction of several characterized individual map elements of the vehicle environment is performed based on the segmented, classified, and transferred bird's-eye view image data, indicated by reference numeral 304. The environmental image data or processed image data is validated and / or supplemented based on standard definition topology data 306 of the vehicle environment. The extracted map elements are combined to create an environment map 310. This is preferably performed by validation using domain knowledge provided by the knowledge graph. Creating a high-resolution environment map 312 (HD map) of the vehicle environment then involves augmenting the environment map 310 using the domain knowledge provided by the knowledge graph 301. This extension is indicated by reference numeral 314 .
[0061] The knowledge graph 301 is here trained based on a training dataset 316 of multiple driving scenes and / or domain knowledge 318. In so doing, the knowledge graph 301 is trained to form spatial relationships between map elements and ensure that only elements that reasonably match each other spatially are used to create and / or supplement the high-resolution environment map 312.
Claims
1. 1. A method for creating a high-resolution environment map for a vehicle having autonomous driving capabilities, comprising: The method comprises: - providing environmental image data of a perception system of the vehicle (S1), the perception system having environmental sensors for capturing the vehicle environment while the vehicle is moving; - providing domain knowledge of the vehicle environment in the form of a trained knowledge graph (S2); - creating a high-resolution environment map of the vehicle environment by supplementing the environmental image data with the provided domain knowledge of the trained knowledge graph (S3); A method comprising:
2. 1. A method for validating and / or supplementing high resolution environmental maps and / or optionally environmental image data for a vehicle having autonomous driving capabilities, comprising: The method comprises: - optionally providing environmental image data (S10) of a perception system of the vehicle, the perception system having environmental sensors for capturing the vehicle environment while the vehicle is moving; - providing a high resolution environmental map of the vehicle environment (S11); - providing domain knowledge of the vehicle environment in the form of a trained knowledge graph (S12); - validating and / or supplementing the high-resolution environment map and / or optionally the environmental image data of the vehicle environment using the provided domain knowledge of the trained knowledge graph (S13); A method comprising:
3. The method according to claim 1 or 2, wherein the environmental image data is validated and / or supplemented based on standard definition topology data of the vehicle environment.
4. The environmental image data is obtained by the following steps: - segmenting image data of the vehicle environment captured by the environmental sensors into elements and classifying said elements into predefined classes that correspond in particular to the ontology of the knowledge graph; - transferring the segmented and classified elements of the vehicle environment into a bird's-eye view; - extracting individual elements of the vehicle environment based on the segmented image data, and / or - combining said extracted elements to create an environment map, in particular by validation with said domain knowledge provided by said knowledge graph. The method according to claim 1 , wherein the water is pretreated by at least one of the following methods:
5. The method of claim 4 , wherein creating the high-resolution environment map of the vehicle environment includes augmenting the environment map with the domain knowledge provided by the knowledge graph.
6. The method according to any one of claims 1 to 5, wherein the environmental sensors include lidar and / or radar sensors and / or cameras.
7. The method of claim 1 , wherein the knowledge graph is trained based on a training dataset of multiple driving scenes and / or domain knowledge.
8. 8. The method of claim 7, wherein the knowledge graph is trained to form spatial relationships between map elements and ensure that only elements that reasonably correspond to each other spatially are used to create and / or supplement the high-resolution environment map.
9. 9. The method according to any one of claims 1 to 8, wherein the knowledge graph comprises domain knowledge about road types and / or lane types and / or lane dividers and / or road boundaries and / or crosswalks and / or stop areas and / or traffic signs and / or traffic lights and / or directional arrows and / or utility poles and / or barriers and / or traffic cones and / or buildings and / or plants and / or debris.
10. An apparatus (100) for creating a high-resolution environment map for a vehicle with autonomous driving capabilities, comprising: The device (100) comprises an evaluation and calculation device, The evaluation and calculation device comprises the steps of: - providing environmental image data to a perception system of the vehicle, the perception system having environmental sensors for capturing the vehicle environment while the vehicle is moving; - providing domain knowledge of the vehicle environment in the form of a trained knowledge graph; - creating a high-resolution environment map of the vehicle environment by supplementing the environmental image data with the provided domain knowledge of the trained knowledge graph; 10. An apparatus configured to perform the steps of:
11. An apparatus (100) for creating a high-resolution environment map for a vehicle with autonomous driving capabilities, comprising: The device (100) comprises an evaluation and calculation device, The evaluation and calculation device comprises the steps of: - optionally providing environmental image data of a perception system of said vehicle, said perception system having an environmental sensor for capturing the vehicle environment while said vehicle is moving; - providing a high resolution environmental map of the vehicle environment; - providing domain knowledge of the vehicle environment in the form of a trained knowledge graph; - validating and / or supplementing the high-resolution environment map and / or optionally the environmental image data of the vehicle environment using the provided domain knowledge of the trained knowledge graph; 10. An apparatus configured to perform the steps of:
12. A control device for a vehicle and / or a robot system and / or an industrial machine having an autonomous driving function, the control device being capable of carrying out the method according to any one of claims 1 to 10.
13. A computer program comprising a program code for performing at least part of the method according to any one of claims 1 to 10, when the computer program is run on a computer.
14. 11. A computer readable data carrier comprising program code of a computer program for performing at least part of a method according to any one of claims 1 to 10, when the computer program is run on a computer.