METHOD AND SYSTEM FOR MAPPING AND LOCALIZING A VEHICLE BASED ON RADAR MEASUREMENTS

Description

[0001] The invention relates to a method for mapping a vehicle environment of at least one vehicle and for locating the at least one vehicle, as well as a system for carrying out such a method. State of the art

[0002] Current autonomous or semi-autonomous vehicles are heavily dependent on provided cartographic information. This cartographic information is used to plan routes and determine the position of the vehicles. The cartographic information or maps are usually available in several layers. A first layer is used, for example, to plan precise lane paths and driving maneuvers. This data enables the autonomous or semi-autonomous vehicles to maintain or adjust a lane. A further layer contains detectable objects that can be detected by on-board sensors and identified by comparing them with the available information in the information layer. Based on the identified objects, the vehicles can determine their relative position to the objects and thus on the map.A further information layer of the map can contain dynamically changing information, such as road conditions, weather conditions, parking information or traffic volume.

[0003] To enable autonomous or semi-autonomous driving functions, this cartographic information must be highly up-to-date and have a high degree of spatial accuracy.

[0004] DE 10 2015 003 666 A1 describes a method for processing measurement data from radar sensors. The measurement data is combined into clusters and stored to generate a digital map. The clusters are formed based on determined spatial clusterings of measurement data within defined cluster radii. The respective clusters can be taken into account when creating the digital map using weighting. If clusters are identified that match previously stored clusters, the previously stored clusters are updated. Disclosure of the invention

[0005] The object underlying the invention can be seen in proposing a method and a system for creating and updating a map and for identifying at least one vehicle position on the map, in which growing measurement errors are at least reduced.

[0006] This object is achieved by means of the respective subject matter of the independent claims. Advantageous embodiments of the invention are the subject matter of the respective dependent subclaims.

[0007] According to one aspect of the invention, a method is provided for mapping a vehicle environment of at least one vehicle and for locating the at least one vehicle. According to the invention, in one step, measurement data of the vehicle environment are determined by at least one radar sensor of the at least one vehicle. The measurement data of the at least one radar sensor are then aggregated and compared with already existing aggregated measurement data. Based on the comparison between the aggregated measurement data and the already existing aggregated measurement data, the aggregated measurement data are optimized by reducing measurement errors, wherein the aggregated measurement data are adapted or interpolated to the already existing aggregated measurement data. A map is created or updated based on the optimized aggregated measurement data.By comparing the measured data with the created or updated map, at least one vehicle is located.

[0008] According to the invention, the aggregated measurement data forms nodes, with connecting paths between the nodes being created and compared for comparison with existing aggregated measurement data. This allows the method to generate nodes and edges analogous to the so-called "graph SLAM" or the simultaneous localization and map creation method and use them to compare and optimize the cartographic information.

[0009] This method allows at least one vehicle to create or update a map with relevant cartographic information. In particular, the accuracy of the map can be increased through repeated measurements. The method can also be used to enlarge and optimize existing maps.

[0010] In the method, radar measurement data is acquired by at least one vehicle and then aggregated. By aggregating the measurement data, coherent point clouds can be identified from the acquired measurement data, which can be used to reduce the total amount of measurement data. This can reduce the amount of measurement data and increase the speed of the process, especially when multiple measurements are being performed or when multiple radar sensors are operating in parallel. Part of the acquired measurement data can also be filtered out by limiting the possible vehicle surroundings.

[0011] The aggregated or condensed point clouds of the measured data can be compared with previously determined point clouds and aligned with each other. As part of this process, for example, distances between the point clouds can be defined and compared in such a way that deviations in the measured data are minimized. The distances between the point clouds or the grouped point clouds can also be determined using odometry based on the radar measurements.

[0012] The error-optimized measurement data or the measurement data from different measurements, available as grouped point clouds, are used to create a map. Further corrections, such as geometric rectification or alignment of the respective measurement data with existing measurement data, can be performed.

[0013] After creating the map or adding measurement data to update the map, the map can be compressed in a further step. This reduces the map's storage requirements, allowing quick access to the map by at least one vehicle.

[0014] Based on the created map, the radar measurement data obtained from at least one vehicle can be used to locate the at least one vehicle. For this purpose, the acquired measurement data is compared with the map to ensure it matches the map. This process can be performed, for example, on an external server unit or in a cloud.

[0015] According to one embodiment of the method, the measurement data from at least one radar sensor are aggregated using a cluster analysis. Cluster analysis can identify similarity structures among the acquired measurement data and the previously stored measurement data. This allows measurement data from multiple measurements to be compressed into a few meaningful measurement data sets, thus compensating for deviations and measurement errors.

[0016] According to one embodiment of the method, the measurement data from at least one radar sensor is filtered. A filter can reduce the amount of measured data acquired for further processing or calculation. In particular, illogical measurement data or data that lie outside a scanning pattern can be deleted. Furthermore, machine learning processes, such as adaptive neural networks, can be used to interpret and filter the measured values.

[0017] According to the invention, at least one node is formed based on at least one measurement data cloud. The measurement data clouds can be aligned and congruent with each other, for example, using an iterative closest point algorithm or any point adjustment algorithm. This allows a distributed measurement data cloud to be condensed, thus increasing the accuracy of further calculations.

[0018] According to one embodiment of the method, the aggregated measurement data is compared with measurement data from at least one second sensor to reduce measurement errors. In addition, at least one further sensor can be used to reduce the errors in the acquired radar measurement data. For example, LIDAR sensors or camera-based sensors can be used to detect prominent objects or geometric shapes in the vehicle's surroundings. These features can be compared with the radar measurement data. This allows erroneous measurement data to be sorted out or optimized, so that a map generated from the measurement values ​​can be more accurate.

[0019] According to one embodiment of the method, the map is updated by overlaying optimized aggregated measurement data. A map can be updated particularly easily using this method if newly determined measurement values ​​overlay the previously stored measurement values ​​or are stored in parallel. Alternatively, existing measurement data can be replaced with new measurement data.

[0020] According to one embodiment of the method, the created map is compressed. This allows the map's storage requirements to be reduced. Particularly when the map is provided by an external server unit, relevant cartographic information can be quickly retrieved by a vehicle, even when connection speeds are impaired depending on the location.

[0021] According to one embodiment of the method, the created map is compressed by clustering, whereby each cluster of the created map is assigned a timestamp and, when the map is updated, corresponding older clusters are replaced by current clusters. The memory requirements of the map created or updated from the determined and compressed measurement data can be reduced in a further step as part of clustering. This can prevent the map from becoming unusable due to its data size and measurement value density. This process can also be used to update existing maps. For this purpose, the measurement data used to update the map can be grouped into clusters beforehand so that the map can be updated or expanded cluster by cluster. In particular, individual clusters of the map can be provided with timestamps.This allows outdated clusters to be deleted when adding current map clusters. In particular, current clusters with a higher weighting can be included for further calculations or route planning.

[0022] According to one embodiment of the method, the compressed map is analyzed to detect objects. The created and subsequently compressed map can be checked for identifiable objects and features. In particular, the measured values ​​determined and used in the form of the map can be analyzed for correlations and recognizable objects or features using a supervised or unsupervised machine learning process or neural networks. This can, for example, extract landmarks, geographical or geometric features from the measured values. This can simplify or accelerate localization of the at least one vehicle. In particular, such a map can also be used by vehicles without radar sensors, provided that the extracted features can be determined and compared by, for example, optical sensors.

[0023] According to one embodiment of the method, the created map is linked to at least one geographical map. This allows the created or updated map to be linked to additional information layers. For example, traffic information or location-dependent weather data can be provided. Conventional GPS data can also be used to increase accuracy and to verify the determined measurement data. This allows a located vehicle to be precisely assigned to a location on the linked maps. The localization of the at least one vehicle can preferably be achieved by comparing measurement points or by filtering the measurement points with the map data. Furthermore, the determined measurement points can be converted in advance as a three-dimensional measurement point cloud into a two-dimensional or 2.5-dimensional point cloud, thus enabling a comparison with a correspondingly created map.

[0024] According to a further aspect of the invention, a system is provided. The system has at least one vehicle with at least one radar sensor for determining measurement data. The system additionally has at least one external processing unit or at least one internal processing unit arranged in the at least one vehicle for creating a map based on the determined measurement data and for locating the at least one vehicle using the determined measurement data. The system is configured to carry out a method according to the invention for mapping a vehicle environment of at least one vehicle and for locating the at least one vehicle.

[0025] The system enables vehicles to create and update maps using radar measurement data, either alone or in combination with at least one external server unit. In particular, this enables the provision of precise and up-to-date cartographic information for autonomous or semi-autonomous driving functions. The measurement data can be processed, for example, on one or more external server units. Multiple external server units can form a cloud service for evaluating and providing the measurement data or cartographic information.

[0026] The method according to the invention can condense or compress radar measurements obtained in a vehicle environment using a "full SLAM" or a "graph SLAM" method, thus generating coherent measurement point clouds. For this purpose, a cluster analysis or sorting of the measurement data can be performed.

[0027] The compressed measurement data can be compared with existing measured values. This can reduce measurement uncertainties and deviations when recognizing measured values ​​or measurement patterns.

[0028] The optimized measurements are then used to generate or update a map. To minimize the storage requirements of the resulting map, the map can be compressed or divided into clusters in a further step.

[0029] The radar measurement data collected by a vehicle can be compared with the measured values ​​stored on a map so that the vehicle can be located on the map.

[0030] In the following, a preferred embodiment of the invention is explained in more detail using a highly simplified schematic representation.

[0031] The Figure 1shows a schematic flow diagram of method 1 for mapping a vehicle environment of at least one vehicle and for locating the at least one vehicle according to a first embodiment.

[0032] In a first step, measurement data of the vehicle surroundings of the at least one vehicle are determined by at least one radar sensor 2. The at least one radar sensor can be arranged in or on the at least one vehicle. The at least one radar sensor can generate radar waves continuously or at defined time intervals and receive reflected radar waves based on a time-of-flight analysis of objects and the vehicle surroundings. For this purpose, the radar sensor has an electronic control and an evaluation unit for controlling the generation of radar waves and for evaluating reflected radar waves. The reflected radar waves received by the radar sensor are determined in the form of measurement data or measurement points and stored at least temporarily.

[0033] In a further step, the measurement data is aggregated 4. This serves in particular to reduce the measurement data density of the determined measurement data and to reduce the storage space required for the measurement data. Thus, the aggregated measurement data can be transmitted, for example, from the at least one vehicle to one or more external server units via a wireless communication connection. The one or more external server units can then take over the further processing steps with a higher provided computing power. Alternatively, the at least one vehicle itself can carry out the processing steps using an internal control unit or processing unit.

[0034] During aggregation of the measurement data 4, only those measurement data or measurement points are retained from the acquired measurement data that are logical and meaningful. In particular, measurement data indicating ghost targets can be deleted in this step 4. For this purpose, a density-based spatial cluster analysis or a so-called k-means algorithm can be applied to the acquired measurement data. During aggregation, the radiation directions of the generated and received radar waves can also be taken into account. This step can be used, for example, to combine the acquired measurement data into measurement point clouds or groups of measurement points, whereby the groups can each depend on a reception angle of the reflected radar waves.

[0035] In a next step, the previously aggregated measurement data is saved. If existing aggregated measurement data is already stored in a memory, a comparison of the current aggregated measurement data with previously saved aggregated measurement data can be performed in this step 6. For this purpose, for example, the distances between the formed groups can be measured or calculated and compared with the distances in the previously stored data. Furthermore, patterns between the different data can be compared. Through such a comparison, the continuously increasing measurement errors of the method can be reduced 8.

[0036] For this purpose, the newly determined and aggregated measurement data are adapted or interpolated to the already stored measurement values.

[0037] Using the measurement data optimized in the previous step 8, a map is then created from the radar-based measurement data 10. If a map has already been created from previous measurements, the optimized measurement data is used to update the map 10.

[0038] The created map can now be used to locate 12 the at least one vehicle. For this purpose, the 2 measured data items determined from at least one vehicle are compared 12 with the measured data 10 stored as a map. A match between the determined measured data 2 and the measured data 10 stored as a map can lead to a position 14 of the at least one vehicle on the map.

[0039] After creating the map from the optimized measurement data 10, the map can be further compressed 16.

[0040] The method 1 thus comprises a part for generating a map M and a part for locating L at least one vehicle using the generated map.

Claims

1. Method (1) for mapping (M) a vehicle environment of at least one vehicle and for locating (L) the at least one vehicle, wherein - measurement data relating to the vehicle environment are determined by at least one radar sensor of the at least one vehicle (2), - the measurement data of the at least one radar sensor are aggregated (4), - the aggregated measurement data (4) are compared with already available aggregated measurement data (6), - the comparison between the aggregated measurement data and the already available aggregated measurement data is taken as a basis for optimizing the aggregated measurement data by reducing measurement errors (8), wherein - the optimized aggregated measurement data are used to create or update a map (10) and - comparison (12) of the determined measurement data with the created map is used to locate the at least one vehicle on the created or updated map (14), characterized in that the aggregated measurement data (4) are matched to the already available aggregated measurement data or interpolated, wherein at least one node is formed on the basis of at least one measurement data cloud, wherein the aggregated measurement data (4) form nodes and comparison with already available aggregated measurement data involves connection paths between the nodes being formed and compared.

2. Method according to Claim 1, wherein the measurement data of the at least one radar sensor are aggregated by way of a cluster analysis (4).

3. Method according to Claim 1 or 2, wherein the measurement data of the at least one radar sensor are filtered.

4. Method according to one of Claims 1 to 3, wherein the aggregated measurement data (4) are compared with measurement data of at least one second sensor to reduce measurement errors (6, 8).

5. Method according to one of Claims 1 to 4, wherein the map is updated by superimposing optimized aggregated measurement data (10).

6. Method according to one of Claims 1 to 5, wherein the created map is compressed (16).

7. Method according to one of Claims 1 to 6, wherein the created map (10) is compressed by clustering (16), wherein each cluster of the created map is assigned a time stamp, and an update of the map results in corresponding older clusters being replaced by current clusters.

8. Method according to Claim 6, wherein the compressed map (16) is analysed to detect objects.

9. Method according to one of Claims 1 to 8, wherein the created map is linked to at least one geographical map.

10. System comprising at least one vehicle having at least one radar sensor for determining measurement data (2) and having at least one external or internal processing unit for creating a map (10) on the basis of the determined measurement data and for locating (L) the at least one vehicle on the basis of the determined measurement data (2), the system being designed to perform a method (1) for mapping (M) a vehicle environment of at least one vehicle and for locating (L) the at least one vehicle according to one of the preceding claims.

11. System according to Claim 10, wherein the system is designed to use the processing unit to perform processing steps of the method according to one of Claims 1 to 9.