EP4172564A1

EP4172564A1 - Determining a starting position of a vehicle for locating

Info

Publication number: EP4172564A1
Application number: EP21732875.6A
Authority: EP
Inventors: Renlin LI; Georg Krause; Timo Nachstedt
Original assignee: Robert Bosch GmbH
Current assignee: Robert Bosch GmbH
Priority date: 2020-06-30
Filing date: 2021-06-14
Publication date: 2023-05-03
Also published as: WO2022002564A1; JP7458514B2; JP2023532729A; US20230204364A1; CN116324338A; DE102020208082A1

Abstract

The invention relates to a method for determining a starting position of a vehicle for locating the vehicle by means of a control device, wherein - measurement data are received from an odometry sensor system and/or a GNSS sensor system of the vehicle, a first position and a range of uncertainty of the first position are determined on the basis of the received measurement data, at least one map section of a feature map having a plurality of stored features is received, wherein the map section has a position and an extension which superimposes the position and the range of uncertainty, measurement data are received from a LIDAR sensor, a radar sensor system and/or a camera sensor system and static features are extracted from the received measurement data, a first starting position of the vehicle is determined by comparing the static features extracted from measurement data to features stored in the map section. The invention furthermore relates to a locating method, a control device, a computer program and a machine-readable storage medium.

Description

description

title

Determining a starting position of a vehicle for localization

The invention relates to a method for determining a starting position of a vehicle for localization of the vehicle. The invention also relates to a method for performing a localization, a control device, a computer program and a machine-readable storage medium.

State of the art

In order to use digital maps with precise lanes, the vehicle must be able to determine its own position with lane accuracy. The position of the vehicle can be determined by localization. The localization of the vehicle position requires a precise initial position or starting position, since the localization is usually based on iterative processes. For example, measurement data from an odometry sensor system are used to determine the vehicle position based on the initial position.

There are already known methods for determining an initial position for vehicle localization, which are based on the reception of GNSS data. Several GNSS positions can be used to determine an initial position. However, the GNSS signals from the corresponding satellites required for this method are not available in many road sections, such as forests or tunnels. Another problem is the insufficient accuracy of the determined GNSS positions. To improve the accuracy of the determined GNSS positions, differential methods can be used, which are costly and not available across the board due to additional hardware.

Disclosure of the invention

The object on which the invention is based can be seen in proposing a precise method for determining an initial position or starting position for an iterative vehicle localization, which can be implemented with inexpensive sensors.

This object is achieved by means of the respective subject matter of the independent claims. Advantageous refinements of the invention are the subject matter of the respective dependent subclaims.

According to one aspect of the invention, a method for determining a starting position of a vehicle for localization of the vehicle by a control device is provided. In particular, the method can be carried out by an initialization module of the control device in the form of hardware and / or software.

In one step, measurement data are received from an odometry sensor system and / or a GNSS sensor system of the vehicle. For this purpose, the odometry sensors and / or GNSS sensors can be connected to the control unit to conduct data.

A first position and an uncertainty range of the first position are determined based on the received measurement data. In a further step, at least one map section of a feature card with a large number of stored features is received. The map section preferably has a position and extent which overlap the first position and the uncertainty region of the first position. The uncertainty area can be designed in the form of an ellipse, a circle or in the form of a polygon. The position can be central or eccentric in the uncertainty area. In a further step, measurement data are received from a LIDAR sensor system, a radar sensor system and / or a camera sensor system and static features are extracted from the received measurement data. Such static features can depict static objects, such as buildings, lane boundaries, lane markings, trees, street signs, lane courses and the like. The static features can be present in the form of a signature or a characteristic measurement data grid.

A starting position of the vehicle is then determined by comparing the static features extracted from the measurement data with the features stored in the map section.

The method can be used to determine an initial position or starting position of the vehicle that is accurate to the lane and serves as the basis for an iterative localization method. The starting position can also be configured as a starting pose with a position and orientation of the vehicle.

By delimiting the map section to a small region around the first position and its uncertainty area, a feature comparison can be carried out with little technical effort. The requirements for the sensor accuracy are low, which means that inexpensive sensors can also be used for the method.

The option of using radar sensors enables the starting position to be determined even in poor weather conditions, such as rain or fog. Furthermore, the starting position can also be precisely determined in areas without a GNSS signal.

The starting position can in particular be used to implement automated or partially automated driving functions. Driving functions of this type can, for example, access lane-accurate navigation and have trajectory planning. The method for determining the starting position is essentially determined by three main steps. In one step, measurement data from the odometry sensors and / or the GNSS sensors of the vehicle are used separately or as a sensor data fusion in order to determine the first position.

Based on the first position, a corresponding map section can be loaded which covers the first position and its uncertainty area. In a third main step, the loaded map section can be used for feature-based localization using measurement data from a LIDAR sensor system, a radar sensor system and / or a camera sensor system.

According to a further aspect of the invention, a control device is provided, the control device being set up to carry out the method. The control device can be, for example, a control device on the vehicle, a control device external to the vehicle, or a server unit external to the vehicle, such as a cloud system, for example.

In particular, the control device can have a localization module and / or an initialization module. As a result, the control device can carry out the method for determining the starting position of a vehicle and / or the method for performing a localization.

In addition, according to one aspect of the invention, a computer program is provided which comprises commands which, when the computer program is executed by a computer or a control device, cause the computer or a control device to execute the method according to the invention. According to a further aspect of the invention, a machine-readable storage medium is provided on which the computer program according to the invention is stored.

According to the BASt standard, the vehicle can be assisted, partially automated, highly automated and / or fully automated or can be operated without a driver.

The vehicle can be, for example, a passenger car, a truck, a robotaxi, and the like. The vehicle is not limited to running on roads. Rather, the vehicle can also be used as a Watercraft, aircraft, such as a transport drone, and the like can be configured.

In a further embodiment, at least one second starting position is determined at a time spaced apart from the first starting position, the first starting position and the at least one second starting position being compared with positions determined from measurement data from the odometry sensor system. A deviation between the at least one second starting position and a position of the vehicle determined by measurement data from the odometry sensor system is preferably calculated and a consistency check is carried out. This consistency check can be implemented as a final step of the method, which ensures reliable results of the initialization. In particular, this consistency check can guarantee the correctness of the starting position in periodic environments, such as a forest section or a motorway, with a large number of identical road boundaries or static features.

The consistency check can be implemented in a technically particularly simple manner if the first of the two successive alignment results or starting positions is propagated by means of measurement data from the odometry sensor system up to the point in time of the second alignment result and then the difference with the actual second alignment result is calculated. If the difference is below a certain threshold, the starting positions can be considered to be consistent.

In an advantageous embodiment, several consistent starting positions can be determined one after the other before the starting position is transferred to the vehicle localization.

According to a further exemplary embodiment, at least one second starting position is determined at a time spaced apart from the first starting position, the first starting position and the at least one second starting position being linked with measurement data from the odometry sensor system to form trajectories. A goodness of fit is determined for each trajectory, where a trajectory with the highest quality of fit is used or all trajectories are discarded. If the starting positions are discarded, the method is carried out again in order to determine a starting position which passes the consistency check. This means that multi-hypothesis tracking can be used. This measure can further increase the precision of the starting position, particularly in periodic environments.

According to a further embodiment, extracted static features from older measurements of the LIDAR sensor system, the radar sensor system and / or the camera sensor system are used to carry out a comparison of the extracted features with the features stored in the map section. The map section can be enlarged according to the distance between a current measurement and the older measurement. This can be achieved by using measurement data from the odometry sensor system. The older measurement data and the current measurement data can preferably have static features which are spatially separated from one another and which, combined with the features stored in the map section, can be compared in order to determine the first starting position. As a result, a larger data cloud can be used to determine the first starting position. This measure can take place, for example, instead of the consistency check.

According to a further exemplary embodiment, the extracted features from the older measurements are coupled with the extracted features from current measurements using measurement data from the odometry sensor system. In this way, the older static features, which were ascertained a few seconds earlier, for example, can be linked to currently ascertained static features over the distance covered by the vehicle. The distance covered can be traced using the odometry sensors. As a result, an enlarged road geometry with corresponding features can be used to resolve ambiguities or ambiguities in periodic surroundings, such as, for example, motorways or forest sections.

According to a further embodiment, an optimization method is used to determine the first position. For example, a data structure based on a sliding graph with measurement data from the odometry sensors and the GNSS sensors of the vehicle received over a period of time, and the first position can be determined therefrom. This measure can eliminate or reduce measurement uncertainties, fluctuations and jitter when determining the first position.

According to a further exemplary embodiment, the measurement data from the odometry sensor system and / or the GNSS sensor system of the vehicle are continuously received. Furthermore, the first position is continuously determined based on the received measurement data. This measure can ensure the availability of the first position for determining the first starting position.

According to a further embodiment, the determined starting position is used to carry out a road clearance service. As a result, the method can be used to release a street or a street section for certain automated or partially automated driver assistance functions. The redundant use of the measurement data from the odometry sensors and the static features can prevent manipulation of the starting position, for example through so-called GNSS spoofing.

According to a further aspect of the invention, a method for performing a localization is provided, a first starting position determined by a method according to the invention for determining a starting position of a vehicle being received as an input variable and / or as a validation variable.

Thanks to the determined starting position, a vehicle can also be localized with low-cost vehicle sensors compared to a digital map. The method for performing the localization can implement a continuous comparison of static features from measurement data from a LIDAR sensor system, a radar sensor system and / or a camera sensor system of the vehicle with the features stored in the digital map. This can be done, for example, by a localization module of the control device. Since the method for performing the localization is designed iteratively, the vehicle can be localized particularly precisely by providing the precise starting position.

In particular, the search area for the feature-based localization can be limited to the at least one map section and the computation requirement can be reduced by the first position.

According to one embodiment, the method for determining the starting position or initial position is carried out in parallel with the method for performing the localization. As a result of this measure, the initialization module of the control device can also run in the background during the localization. The output of the initialization module can be used as an additional status check of the localizer output. If a difference between the starting position and the vehicle position determined by the localization is determined which exceeds a limit value, a new starting position can be output for the localization module and the localization of the vehicle can be initiated again.

In the following, preferred exemplary embodiments of the invention are explained in more detail on the basis of greatly simplified schematic representations. Show here

1 shows a schematic plan view of a roadway to illustrate a determination of a first position of a vehicle,

FIG. 2 shows a schematic plan view of a roadway from FIG. 1 and a map section,

3, 4 are schematic diagrams to illustrate a consistency check,

FIGS. 1 to 4 show schematic representations to illustrate a method for determining a first starting position A of a vehicle 2 for localization of the vehicle 2 by a control device 4. The vehicle 2 has an odometry sensor system and / or a GNSS sensor system 6 and an additional sensor system 8 for feature-based localization. The additional sensor system 8 can be configured, for example, as a LIDAR sensor system, a radar sensor system and / or a camera sensor system.

FIG. 1 shows a schematic top view of a roadway 10 to illustrate a determination of a first position P of the vehicle 2. The vehicle 2 travels on the roadway 10 in the direction of travel F. In the exemplary embodiment shown, measurement data are recorded by the odometry sensors and the GNSS sensors while driving 6 collected. A plurality of measurement data 12 from the odometry sensor system and the GNSS sensor system 6 collected one after the other are stored in order to determine the first position P. The measurement data 12 of the odometry sensor system and the GNSS sensor system 6 can optionally be smoothed so that optimized measurement data 14 result from the measurement data 12 determined. The determined measurement data 12 can be optimized, for example, by means of moving averages.

The determined measurement data 12 can have a maximum number, so that old measurement data are automatically deleted or overwritten by more recent measurement data. The first position P can be configured as the last or most recent measurement of the odometry sensor system and the GNSS sensor system 6 after optimization or smoothing.

FIG. 2 shows a schematic top view of a roadway 10 from FIG. 1 and of a map section 16. The map section 16 is part of a feature map and has a multiplicity of features 18. In the exemplary embodiment shown, the feature map is a radar-specific feature map. The map section 16 has a position and an extent which overlays or covers the first position P and an optional uncertainty region of the first position P.

Furthermore, extracted static features 20 from current measurements and static features 22 from older measurements of the LIDAR sensor system, the radar sensor system and / or the camera sensor system 8 are taken into account, so that an enlarged map section 16 is used to carry out a comparison of extracted static features 20, 22 with features 18 stored in the map section 16.

The radar-specific feature map and the map section 16 are stored in the form of map sections 16 which depict the topology of the road network or the roadway 10. The at least one map section 16 can be transformed into a coordinate system of the vehicle 2, which is not shown in FIG. 2 for the sake of clarity. Furthermore, the features 18 of the map section 16 can be compared with the extracted static features 20, 22 and aligned with one another for a feature-based localization. Such an alignment can be realized with a cost function and an optimization algorithm for the cost function.

FIG. 3 and FIG. 4 show schematic diagrams to illustrate a consistency check. FIG. 3 shows a technically simplified consistency check. A plurality of second starting positions A2, A3 spaced apart in time from the first starting position A are determined.

In parallel to the starting positions A, A2, A3, positions P, P2, P3 are determined by means of the odometry sensor system 6 and compared with the starting positions A, A2, A3. For this purpose, a difference D or a distance between the positions P, P2, P3 and the starting positions A, A2 can be calculated. The consistency check is successful if the difference D is below a predefined threshold value or limit value.

In the illustrated embodiment, the first two starting positions A, A2 are consistent or correct. The last starting position A3 has too great a deviation D from position P3 and is not consistent. Here, the method for determining the starting position A of the vehicle 2 can be carried out again and subjected to a consistency check. The consistency check can preferably have a plurality of successfully checked starting positions A, A2, A3 before the last checked starting position A3 is released for localization of the vehicle 2. A technically more complex consistency check is illustrated in FIG. 4, which is based on multi-hypothesis tracking. A plurality of starting positions A, A2, A3 with a best match within the map section 16 are taken into account. These starting positions A, A2, A3 are coupled with matches from the last alignment of the features 18, 20, 22. The measurement data 12 of the odometry sensor system 6 are used to connect the starting positions A, A2, A3. Each of the starting positions A, A2, A3 can have parallel starting positions AP within the map section 16, in which the extracted features 20,

22 match the features 18 stored in the map section 16. Such results of the feature-based localization can occur in periodic environments such as, for example, motorway sections. The measurement data 12 of the odometry sensor system 6 are compared in the form of trajectories with the respective starting positions A, A2, A3, AP, the starting positions A, A2, A3, AP with the highest degree of adaptation being used for the further localization of the vehicle 2.

Claims

Expectations

1. A method for determining a starting position (A) of a vehicle (2) for localization of the vehicle (2) by a control unit (4), wherein

Measurement data (12) are received from an odometry sensor system and / or a GNSS sensor system (6) of the vehicle (2), based on the received measurement data (12) a first position (P) and an uncertainty range of the first position (P) are determined , at least one map section (16) of a feature map with a large number of stored features (18) is received, the map section (16) having a position and extent which superimposes the first position (P) and the uncertainty area, measurement data from a LIDAR Sensor system, a radar sensor system and / or a camera sensor system (8) are received and static features (20) are extracted from the received measurement data, a first starting position (A) of the vehicle (2) by comparing the static features (20) extracted from the measurement data in the map section (16) stored features (18) is determined.

2. The method according to claim 1, wherein at least one second starting position (A2, A3) is determined at a time spaced apart from the first starting position (A), the first starting position (A) and the at least one second starting position (A2, A3) with from measurement data (12) positions (P, P2, P3) determined by the odometry sensor system (6) are compared, a deviation (D) between the at least one second starting position (A2, A3) and one determined by measurement data (12) from the odometry sensor system (6) Position (P2, P3) of the vehicle (2) is calculated and a consistency check is carried out.

3. The method according to claim 1, wherein at least one second starting position (A2, A3, AP) is determined at a time spaced apart from the first starting position (A), the first starting position (A) and the at least one second starting position (A2, A3, AP) ) are linked with measurement data (12) of the odometry sensor system (6) to form trajectories, a quality of fit being determined for each trajectory, a trajectory with the highest quality of fit being used or all trajectories being discarded.

4. The method according to claim 1, wherein extracted static features (22) from older measurements of the LIDAR sensors, the radar sensors and / or the camera sensors (8) for performing a comparison of the extracted features (22) with stored in the map section (16) Features (18) are used.

5. The method according to claim 4, wherein the extracted static features (22) from the older measurements by measurement data (12) of the odometry sensor system (6) are coupled with the extracted static features (20) from current measurements.

6. The method according to any one of claims 1 to 5, wherein an optimization method is used to determine the first position (P).

7. The method according to any one of claims 1 to 6, wherein the measurement data (12) of the odometry sensors and / or the GNSS sensors (6) of the vehicle (2) are continuously received and based on the received measurement data (12) continuously the first position (P) is determined.

8. The method according to any one of claims 1 to 7, wherein the determined starting position (A) is used to carry out a road clearance service.

9. A method for performing a localization, wherein a first starting position (A) determined by a method according to one of the preceding claims is received as an input variable and / or as a validation variable.

10. The method according to claim 9, wherein the method for determining the starting position according to one of claims 1 to 8 is carried out in parallel with the method for performing the localization according to claim 9.

11. Control device (4), wherein the control device (4) is set up to carry out the method according to one of claims 1 to 8 and / or the method according to one of claims 9 or 10.

12. Computer program, which comprises instructions that are used in the execution of the

Computer program by a computer or a control device (4) cause the latter to carry out the method according to one of claims 1 to 8 and / or the method according to one of claims 9 or 10.

13. Machine-readable storage medium on which the computer program according to claim 12 is stored.