Method for extracting features from cloud-based collected measurement data - Patents.com

Abstract

To provide a method for determining or extracting feature from measured data, the method being implementable with less technical cost.SOLUTION: A method 1 for ascertaining features for digital maps, in particular for digital HAD maps by control equipment is disclosed. In this case, ascertained measured data 22 of a sensor of at least one vehicle are received or retrieved. The measured data 22 are superimposed on each other according to their geographic coordinates. The superimposed measured data 22 are linked to each other and optimized with respect to errors detected with superimposition. The measured data 22 are clustered and features are extracted 26 from the formed clusters. The extracted 26 features are provided for updating or creating the digital maps. The control equipment, a computer program, and a machine-readable memory medium are also described.SELECTED DRAWING: Figure 5

Description

[Technical field]

[0001] The present invention relates to a method for extracting features from crowd-based collected measurement data, in particular for determining features for a digital map, in particular for a digital HAD map, by a control device. The present invention further relates to a control device, a computer program, and a machine-readable memory medium. [Background technology]

[0002] Automatically operable vehicle Autonomous driving functions are currently guided by sophisticated maps, usually implemented as digital Highly Automated Driving (HAD) maps, which have multiple levels.

[0003] A map may for example have a planning level and a location level: the planning level is used to perform trajectory planning and presents road extents and road geometry.

[0004] Using the location level, vehicle is determined by the sensor vehicle The surrounding features can be matched with the virtual features in the localization level and the planning level. vehicle The location of the feature can be determined, for example, by a radar sensor.

[0005] Until now, special measurements vehicle Methods are known in which a HAD map with a localization level and a planning level is created from the measurement data of the HAD map. To create and store the features of the localization level, measurement data of different types of sensors, for example radar sensors and camera sensors, are usually compared with each other. Such measurement methods can only meet the requirements of the HAD map with regard to updateability and accuracy at a high cost. Summary of the Invention [Problem to be solved by the invention]

[0006] The problem on which the present invention is based can be found in proposing a method for determining or extracting features from measurement data which can be carried out at lower technical cost. [Means for solving the problem]

[0007] This problem is solved by the subject matter of the independent claims. Advantageous configurations of the invention are the subject matter of the respective dependent dependent claims.

[0008] According to one aspect of the invention, a method is provided for determining features for a digital map, in particular for a digital HAD map, by a control device. In one step, at least one vehicle The determined measurement data of at least one sensor of the same class is received or called. vehicle Preferably, the measurement data of the at least one sensor may be referenced. The at least one sensor may be a radar sensor, a LIDAR sensor, an ultrasonic sensor, a camera sensor, and the like. Furthermore, a combination of different or the same sensors may be used.

[0009] Alternatively or additionally, for the method, vehicle In this case, only the measurement data of a plurality of sensors may be available.

[0010] Depending on the form of the control device, the measurement data may already be present in the memory and can be called up, or one or more vehicle can be called from vehicle are to be understood in particular as mobile units, which may also be formed as robots, ships, aircraft and the like. vehicle The concept is not limited to automobiles.

[0011] In a further step, different vehicleThe measurement data determined by are superimposed on one another according to their geographic coordinates. This step can be carried out by rough superimposition of the measurement data using GPS information.

[0012] GPS information, Vehicle Location and Vehicle Preferably, the measurement data can include the orientation. The alignment of the measurement data can be inaccurate due to noise and systematic errors. For this reason, the superimposed measurement data are linked to each other and optimized with respect to errors. By optimization, the errors identified by the superimposition can be eliminated and the spatial offset of the respective measurement data sets from each other can be minimized.

[0013] In a further step, the measurement data of the different measurement data sets are grouped into object clusters and features are extracted from each object cluster. By clustering, measurement data crowds or dense measurement data clouds are detected and groups or clusters are formed. Each cluster can then be extracted in the form of a feature.

[0014] The extracted features may then be used or provided in creating or updating the planning and / or localization levels of the HAD map.

[0015] Optimization and clustering of the measurement data set can make feature extraction more efficient and matching with measurement data of further sensor classes, such as camera sensors, can be omitted. The method can in particular determine and extract features using only one sensor class.

[0016] This feature extraction is non-overlapping and distinct. vehicleAn object cluster or a measurement data cluster representing a plurality of objects detected by the method can be reduced to a virtual feature or a virtual object for use in the HAD map. This step can be performed, for example, by a cluster algorithm. The feature can be, inter alia, an object. The feature can be formed, for example, as a traffic light, a sign, a pole, a tree, and the like.

[0017] The optimization of the measurement data set can take place, for example, by association of the measurement data with adjacent virtual objects at the planning level, for which attributes or features of the planning level can be called up in parallel to the measurement data, and this step can result in a synchronization of the received measurement data with existing features of the planning level.

[0018] According to a further aspect of the invention, a control device is provided, the control device being adapted to carry out the method. The control device may e.g. vehicle External control device or vehicle It may be an external server unit, for example a cloud system. The control device is preferably capable of receiving measurement data of at least one sensor and / or measurement data of a plurality of sensors.

[0019] Moreover, according to an aspect of the invention there is provided a computer program comprising instructions which, when executed by a computer or control device, cause the computer or control device to carry out the method according to the invention. According to a further aspect of the invention there is provided a machine-readable memory medium having stored thereon a computer program according to the invention.

[0020] At least one of these vehicle The system may be assisted in accordance with BASt (German Road Traffic Institute) standards, partially automated, highly automated, and / or fully automated or capable of operating without a driver. vehiclemay be, inter alia, a mobile unit formed as a vehicle, a robot, a drone, a ship, a rail vehicle, a robotaxi, an industrial robot, a commercial vehicle, a bus, an airplane, a helicopter, and the like.

[0021] The method and control device according to the invention can simplify and accelerate the complex processing steps for extracting features for map creation or map updating. vehicle The measurement data collected thereby can be optimally utilized and used at planning level.

[0022] According to one exemplary embodiment, at least two vehicle The measurement data or measurement data sets determined by the sensors are associated with and geolocated at least one sensor-specific localization map. By such a procedure, a particularly precise overlay of the respective measurement data sets can be performed, which results in a particularly accurate feature extraction of attributes or features for the planning level.

[0023] According to a further embodiment, at least two vehicle The measurement data determined by the sensors of the map are geographically aligned by means of preset alignment attributes. For this purpose, dedicated alignment attributes can be provided for the accurate alignment of the measurement data sets and can be used for the alignment and association of the measurement data sets. The alignment attributes can be provided in addition to the features of the planning level in order to enable the measurement data to be aligned or corrected to existing features of the planning level. This allows the collection of data for the planning level of the map to be separated from the alignment step of the respective data.

[0024] According to a further exemplary embodiment, at least two vehicle and / or for work vehicle The measurement data determined by the sensors of the (fleet) Vehicle It is received or retrieved by an external control device and processed to extract features. VehicleThe external control device can in this case be formed as a server unit or a cloud. vehicle can send the determined measurement data or measurement data sets to the control device continuously or at defined time intervals. vehicle A communication connection based on, for example, WLAN, GSM, UMTS, LTE, 5G and similar transmission standards can be provided between the control device and the measurement data. This measure allows for crowdsourcing of the measurement data, with the control device acting as a central unit.

[0025] According to a further embodiment, the determined measurement data comprises: vehicle The measurement data is received by a control device on the other side and processed to extract features, whereby the processing of the measurement data is at least in part performed by each of the control devices that collect the measurement data. vehicle In this case, the results of the feature extraction can likewise be transmitted to a central server unit or cloud and stored there. In particular, the optimization and extraction steps can be performed in a central server unit or cloud, so as to take advantage of relatively high and stable computing power. Vehicle With this measure, the adaptation of the algorithms for measurement data association and optimization can also be performed centrally.

[0026] According to a further exemplary embodiment, the features are extracted in the form of objects and transmitted to the map planning level. The features can be formed as traffic lights, signs, poles, trees, and the like, among others. The objects are preferably identified and extracted from the clusters of measurement data, which also allows for a location determination that is inclusive of the sensor class.

[0027] According to a further embodiment, the measurement data is determined by a radar sensor, a LIDAR sensor, an acoustic sensor or a camera sensor, so that the method is not limited to a specific sensor class, in particular features can be isolated from the measurement data of different sensor classes or can be extracted in relation to multiple sensor classes.

[0028] In the following, preferred exemplary embodiments of the invention are explained in more detail on the basis of highly simplified schematic drawings. [Brief description of the drawings]

[0029] [Figure 1] FIG. 1 is a flow diagram to illustrate a method according to one embodiment in comparison with a map-making run. [Diagram 2] FIG. 2 shows steps to illustrate a method according to one embodiment. [Diagram 3] FIG. 2 shows steps to illustrate a method according to one embodiment. [Figure 4] FIG. 2 shows steps to illustrate a method according to one embodiment. [Diagram 5] FIG. 11 is a flow diagram illustrating a method according to a further embodiment. [Figure 6] FIG. 1 is a flow diagram illustrating a method according to one embodiment of using alignment attributes. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0030] In FIG. 1, a flow diagram is shown to illustrate a method 1 according to one embodiment in comparison with a conventional mapping run 2.

[0031] In the conventional map-making run 2, the measurement data 6 is generated by a special measurement vehicle 4 and are then processed to extract features 8 for the HAD map. Here, a conventional map-making run 2 can be implemented in addition to the method 1 according to the invention.

[0032] In the method 1, the measurement data or measurement data set 10, 11 is a plurality of vehicle The measurement data 10, 11 are collected by, for example, Vehicle The measurement data 10, 11 may be pre-processed 14, 15 internally. vehicle The data may be collected by radar and / or LIDAR sensors on the side.

[0033] The pre-processed measurement data 14, 15 are superimposed and correlated in a further step 16. The alignment or superimposition of the measurement data 10, 11 may be inaccurate due to noise and systematic errors. vehicle The measurement data 10, 11 determined by 12, 13 are offset from each other.

[0034] Here, the preprocessed measurement data14,15 Vehicle The signal is then transmitted to an external control device 3 for further processing. vehicle Between 12, 13 and the control device 3, a communication connection 7, for example based on WLAN, GSM, UMTS, LTE, 5G and similar transmission standards, can be provided. With this measure, crowdsourcing can be realized, in which the control device 3 acts as a central unit for the measurement data 10, 11, 14, 15. The control device 3 is exemplarily formed as a cloud, from which the measurement data 10, 11, 14, 15 can be collected by crowdsourcing and used to extract features 20.

[0035] The extracted features 20 may be stored, for example, in a machine-readable memory medium 5 .

[0036] In a further step 18, an optimization of the superimposed measurement data 16 takes place, whereby for example errors and offsets can be removed, so that a deviation A of the superimposed measurement data 16 can be reduced. One or more features can then be extracted 20 from the corrected measurement data 18.

[0037] 2, 3 and 4 exemplarily show steps for illustrating the method 1 according to one embodiment. The feature 20 to be determined or extracted is exemplarily formed as a traffic light. The traffic light is not shown. vehicle The detected signals are detected by radar sensors 12 , 13 and evaluated by the control device 3 in the form of measurement data 10 , 11 .

[0038] In Figure 2, different vehicleIn order to clarify the deviation A of the measurement data 16 determined by 12 and 13, vehicle 12, 13 are shown together with the superimposed measurement data 16. Due to poor registration, multiple traffic lights would be extracted as features if a planning map were created.

[0039] 3 shows a further step of the method 1, in which the superimposed measurement data 16 are corrected and optimized. This step reduces the deviation A of the respective measurement data 16 and vehicle The traffic light positions determined by 12, 13 are dense. In this respect, a traffic light which actually exists alone is still represented by several traffic lights determined by the sensors.

[0040] 4 shows in a further step the extraction of a feature 20 from the dense cluster of measurement data representing traffic lights, by which a cluster algorithm is used to form a feature 20 of a volume of measurement data, thus reducing the volume of measurement data to only one object.

[0041] The features 20, exemplarily embodied as traffic lights, can then be used for the mapping of the planning and / or localization levels of the HAD map. Moreover, the method also allows for the extraction of multiple features simultaneously or one after the other. For this purpose, measurement data crowds are identified, clustered and extracted as features 20. Essentially all objects can be used as features 20. In particular, trees, support structures, such as lampposts or signposts, traffic lights, signs and the like can be used as features 20.

[0042] 5 shows a flow diagram to illustrate a method 1 according to a further embodiment. Crowdsourced measurement data 22 are received by a control device 3. For this purpose, a number of vehicle These measurement data sets are received. vehiclecan collect measurement data sets during normal driving and send them to the control device 3 or to the cloud.

[0043] The measurement data are then aligned 24 on the basis of the localization map or localization level L of the HAD map. vehicle The measurement data sets determined by can be registered and superimposed with respect to one another.

[0044] The alignment 24 of the measurement data based on the localization level L may then be used for clustering and extraction 26 of features 20 for use in the planning level P.

[0045] In Fig. 6 a flow diagram is shown to illustrate method 1 according to a further embodiment. The use of alignment attributes is highlighted.

[0046] In one step, crowdsourced measurement data 22, e.g. vehicle or vehicle Fleet radar measurement data is received. In parallel with this, planning level P data 30 may be received along with alignment attributes 32.

[0047] The data and alignment attributes 32 of plan level P are similarly vehicle can be determined by measurement or can be provided, for example, by a mapping company, to facilitate frequent and automatic updates of features 20 in a map.

[0048] Based on the collected measurement data 22 and the alignment attributes 30, the collected measurement data 22 is aligned based on the localization level L. The aligned measurement data is then optimized and clustered. Afterwards, features 20 can be extracted 26 from the measurement data.

Claims

1. A method (1) for determining features (20) for a digital map by a control device (3), comprising: receiving a plurality of determined measurement data (10, 11, 22) of at least one sensor of at least one vehicle (12, 13); - superimposing said plurality of measurement data (10, 11, 22) on one another according to their geographic coordinates; - correlating the superimposed measurement data (10, 11, 22) with each other and correcting errors, which are spatial deviations obtained by the superposition between the measurement data (10, 11, 22), - clustering the corrected, superimposed measurement data (10, 11, 22) and identifying objects from the clusters (18) formed and extracting features (20) in the form of said objects; - submitting said extracted features (20) for updating or creating a digital map; The method (1) includes:

2. The method of claim 1, wherein the features (20) are determined for a digital HAD map.

3. 2. The method according to claim 1, wherein the plurality of measurement data (10, 11, 22) determined by the sensors carried by at least two of the vehicles (12, 13) are associated with and geographically aligned with at least one sensor-specific location map (L).

4. the plurality of measurement data (10, 11, 22) determined by the sensors of at least two of the vehicles (12, 13) are geographically aligned according to a preset alignment attribute (32); said alignment attribute (32) enabling said plurality of measurement data (10, 11, 22) to be aligned to existing features of a map plan level (L); The planning level (L) of the map is used to carry out the trajectory planning and indicates the road extent and road geometry.

4. The method according to any one of claims 1 to 3.

5. 5. The method according to claim 1, wherein a plurality of measurement data (10, 11, 22) determined by sensors of at least two of the vehicles (12, 13) and / or work vehicles are received by a control device (3) external to the vehicle and processed to extract features (20).

6. 5. The method according to claim 1, wherein the determined measurement data (10, 11, 22) are received by a vehicle-side control device (3) and processed to extract the features (20).

7. 5. The method of claim 4, wherein said features (20) are transmitted to a planning level (P) of said map.

8. The method according to any one of claims 1 to 7, wherein the plurality of measurement data (10, 11, 22) is determined by a radar sensor, a LIDAR sensor, an acoustic sensor or a camera sensor.

9. A control device (3) adapted to carry out the method according to any one of claims 1 to 8.

10. A computer program comprising commands which, when executed by a computer or control device (3), cause said computer or control device (3) to carry out a method according to any one of claims 1 to 8.

11. A computer readable memory medium (5) having stored thereon a computer program according to claim 10.

Images