DE102021004535A1 - Method for calibrating an odometry device of a motor vehicle, control device for carrying out such a method, calibration device with such a control device and motor vehicle with such a calibration device - Google Patents

Abstract

The invention relates to a method for calibrating an odometry device (5) of a motor vehicle (1) second point in time a second depth recording (11.2) of the surroundings of the motor vehicle (1) is created by means of the sensor (7), with a motion vector (15) from the odometry device (5) for a movement of the motor vehicle (1) between the first point in time and the second point in time is provided, wherein - the first depth recording (11.1), the second depth recording (11.2) and the motion vector (15) are compared, wherein - based on the comparison (19) at least one parameter (17) of the odometry device (5) is assessed and calibrated.

Description

The invention relates to a method for calibrating an odometry device of a motor vehicle, a control device for carrying out such a method, a calibration device with such a control device and a motor vehicle with such a calibration device.

A motor vehicle that drives autonomously, in particular, has a plurality of assistance systems that create a representation of the surroundings of the motor vehicle. The representation of the surroundings of the motor vehicle is used to ensure safe operation of the motor vehicle. Reliable motion estimation and position determination of the motor vehicle by means of odometry data are necessary for a correct representation of the environment. An odometry device providing the odometry data has to be calibrated in a complex manner using a plurality of parameters. In addition, it is possible for the majority of parameters to change dynamically, in particular when a tire is changed, when there are signs of wear or as a result of environmental influences.

Methods for calibrating an odometry device are known in which a plurality of driving maneuvers, in particular circular driving, incline driving, figure driving and slalom driving, taking into account a speed, a driving direction and a curve radius, are carried out and analyzed by means of an external sensor system. These methods, in particular for the initial calibration of the odometry device, are time-consuming.

In addition to the, in particular, initial calibration of the odometry device, it is preferably necessary to calibrate the odometry device dynamically, in particular while the odometry device is in operation.

For particularly dynamic calibration of the odometry device, camera-based methods are known in which a first wheel circumference is estimated by means of a camera and is used as one of the plurality of parameters. After a particularly predefined period, a second wheel circumference is estimated and compared with the first wheel circumference. If a difference between the first wheel circumference and the second wheel circumference exceeds an in particular predefined threshold value, the odometry device is calibrated, in particular in accordance with the second wheel circumference. A disadvantage of this method is that, in particular, the accuracy of the optical estimation of the wheel circumference is heavily dependent on weather and / or light conditions, and thus a robust and reliable calibration of the odometry device cannot be guaranteed.

The invention is therefore based on the object of creating a method for calibrating an odometry device of a motor vehicle, a control device for carrying out such a method, a calibration device with such a control device and a motor vehicle with such a calibration device, the disadvantages mentioned being at least partially eliminated, preferably are avoided.

The object is achieved in that the present technical teaching is provided, in particular the teaching of the independent claims and the embodiments disclosed in the dependent claims and the description.

The object is achieved in particular by creating a method for calibrating an odometry device of a motor vehicle. At a first point in time, a first depth recording of the surroundings of the motor vehicle is created by means of a sensor. Furthermore, at a second point in time, a second depth recording of the surroundings of the motor vehicle is created by means of the sensor. The odometry device provides a motion vector for a movement of the motor vehicle between the first point in time and the second point in time. The first depth recording, the second depth recording and the movement vector are then compared, with at least one parameter of the odometry device being evaluated and calibrated based on the comparison.

A sensor which the motor vehicle has can advantageously be used as the one sensor, so that external sensor systems and / or expensive reference sensors, which in particular are used exclusively for calibration, can be saved. Furthermore, the calibration is advantageously independent of the surroundings of the motor vehicle and / or environmental influences. In addition, a particularly time-consuming and / or financial outlay for calibrating the odometry device, in particular after a tire change, is advantageously reduced in comparison to the known methods. In addition, the method is suitable for calibrating the odometry device in a low speed range of the motor vehicle, in particular at a speed of less than 30 km / h.

A sensor selected from a group consisting of an SFM vision sensor, a USS sensor, a radar sensor and a lidar sensor is preferably used as the one sensor.

In the context of the present technical teaching, an SFM vision sensor is an optical sensor which calculates 3D information by means of a plurality of images recorded in particular with a time delay and a “structure from motion” method (SFM).

In the context of the present technical teaching, it is possible to use the motion vector to describe a movement with three degrees of freedom of rotation and three degrees of translational freedom. In this respect, the motion vector is a coordinate transformation with three degrees of freedom of rotation and three degrees of translational freedom.

The comparison of the first depth recording, the second depth recording and the movement vector is preferably carried out by means of a method for determining an index of a structural similarity, in particular by means of an SSIM method.

In the context of the present technical teaching, in a method for determining an index of a structural similarity, in particular an SSIM method, the similarity between two recordings, in particular images, is measured. Such a procedure is used in "Image quality assessment: from error visibility to structural similarity" by Z. Wang et. al., IEEE Transactions on Image Processing, Volume 13, No. 4, April 2004, 600-612 described.

In particular, the at least one parameter of the odometry device is selected from a group consisting of a wheel circumference, an abrasion coefficient, a wheelbase, a weather situation and a temperature, in particular an ambient temperature.

According to a development of the invention, it is provided that the first depth recording has a first point cloud, and that the second depth recording has a second point cloud. The first depth record preferably consists of the first point cloud and the second depth record preferably consists of the second point cloud.

In a preferred embodiment of the method, at least one point cloud is selected from the first point cloud and the second point cloud, preferably both point clouds, by means of a method for corner and edge detection according to the publication “Edge and Corner Detection for Unorganized 3D Point Clouds with Application to Robotic Welding ”by SM Ahmed et. al. - https://arxiv.org/ftp/arxiv/papers/1809/1809.10468.pdf - analyzed. The comparison is particularly preferably carried out based on a detected corner and / or edge.

In a further preferred embodiment of the method, at least one point cloud is selected from the first point cloud and the second point cloud, preferably both point clouds, by means of a method for corner detection and feature line tracing according to the publication "Edge Detection and Feature Line Tracing in 3D Point Clouds by Analyzing Geometric Properties of Neighborhoods ”by H. Ni et. al., Remote Sens. 2016, 8 (9), 710 - https://doi.org/10.3390/rs8090710 - analyzed. The comparison is particularly preferably carried out based on a detected corner and / or a feature line.

In a further preferred embodiment of the method, at least one point cloud is selected from the first point cloud and the second point cloud, preferably both point clouds, by means of a method for contour detection according to the publication “Contour detection in unstructured 3D point clouds” by T. Hackel et. al., Photogrammetry and Remote Sensing, ETH Zurich - https://ethz.ch/content/dam/ethz/special-interest/baug/igp/photogrammetry-remotesensing-dam/documents/pdf/timo-jan-cvpr2016.pdf - analyzed. The comparison is particularly preferably carried out based on a detected contour.

In a further preferred embodiment of the method, at least one point cloud is selected from the first point cloud and the second point cloud, preferably both point clouds, by means of a method for feature extraction according to the publication “3D Feature Point Extraction from Lidar Data using a Neural Network” by Y. Feng et. al., The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Volume XLI-B1, 2016 and XXIII ISPRS Congress, 12-19 July 2016 , Prague, Czech Republic - https://www.repo.unihannover.de/bitstream/handle/123456789/714/isprs-archives-XLI-B1-563-2016.pdf?sequence=1&isAllowed=y - analyzed. The comparison is particularly preferably carried out based on an extracted feature.

In a further particularly preferred embodiment of the method, a first surface spanning the first point cloud is determined. Alternatively or additionally, a second surface spanning the second point cloud is preferably determined. In particular, a predetermined number of normal vectors of the first surface and / or the second surface is determined, the predetermined number of normal vectors being evaluated by means of a histogram. In particular, a unimodal distribution of the histogram indicates a flat surface, in particular a wall. Furthermore, in particular a biomodal distribution of the histogram points to an edge, in particular a transition from one Floor to a wall. Furthermore, a histogram with three local maxima indicates a corner point. The comparison is particularly preferably carried out based on a detected edge and / or a detected corner.

According to a development of the invention it is provided that a similarity value is determined for the first depth recording, the second depth recording and the motion vector. The similarity value is compared with a threshold value, the at least one parameter of the odometry device being evaluated and calibrated based on the comparison of the similarity value with the threshold value.

This advantageously prevents the odometry device from being calibrated due to noise in at least one depth recording selected from the first depth recording and the second depth recording.

According to a development of the invention it is provided that at least one first point of the first point cloud is transformed, in particular rotated and / or translated, by means of the motion vector, whereby at least one third point is obtained. The at least one third point is compared with a second point of the second point cloud associated with the at least one first point, the at least one parameter of the odometry device being evaluated and calibrated based on the comparison.

A distance between the third point and the second point, in particular in a common coordinate system, is preferably determined as a comparison between the third point and the second point. If the distance in particular exceeds a predetermined threshold distance, at least one parameter of the odometry device is calibrated.

The comparison between the third point and the second point is particularly preferably carried out by means of a method for determining an index of a structural similarity of points, in particular by means of a PointSSIM method.

In particular, the first point and the second point assigned to the first point represent an identical feature of the surroundings of the motor vehicle.

If the motion vector of the odometry device correctly maps the motion of the motor vehicle, in particular the third point is identical to the second point. If the movement vector of the odometry device does not correctly map the movement of the motor vehicle, in particular the third point is different from the second point.

In particular, a third point cloud is obtained from the third point, preferably from a plurality of third points.

If the motion vector of the odometry device correctly maps the movement of the motor vehicle, in particular a first first plurality of points of the third point cloud are identical to a first second plurality of points of the second point cloud and a second first plurality of points of the third point cloud are different from a second second plurality at points of the second point cloud. The first first plurality is preferably larger than the second first plurality and / or the first second plurality is preferably larger than the second second plurality.

If the motion vector of the odometry device does not correctly map the movement of the motor vehicle, in particular the first first plurality of points of the third point cloud are identical to the first second plurality of points of the second point cloud and the second first plurality of points of the third point cloud are different from the second second Majority of points in the second point cloud. The second first plurality is preferably larger than the first first plurality and / or the second second plurality is preferably larger than the first second plurality.

According to a development of the invention it is provided that at least one second point of the second point cloud is transformed by means of the motion vector, in particular rotated and / or translated, whereby at least one fourth point is obtained. The at least one fourth point is compared with a first point of the first point cloud assigned to the at least one second point, the at least one parameter of the odometry device being evaluated and calibrated based on the comparison.

In particular, a fourth point cloud is obtained from the fourth point, preferably from a plurality of fourth points.

In particular, a first transformation between the first point cloud and the third point cloud and a second transformation between the second point cloud and the fourth point cloud are designed in such a way that the third point cloud is converted into the first point cloud by means of the second transformation and that the fourth point cloud is converted into the first point cloud by means of the first Transformation is transferred to the second point cloud.

According to a development of the invention it is provided that a first maximum of the first surface spanning the first point cloud is determined as the at least one first point. Furthermore, as the second point assigned to the at least one first point, a second maximum, assigned to the at least one first maximum, of the second surface spanning the second point cloud is determined.

It is advantageously possible in a simple manner to track the first maximum and thus to determine the second maximum associated with the first maximum in the second point cloud.

In particular, the first maximum is a local maximum. Alternatively, the first maximum is preferably a global maximum.

In particular, the second maximum is a local maximum. Alternatively, the second maximum is preferably a global maximum.

The first maximum is preferably transformed by means of the motion vector into a third maximum of a third surface spanning the third point cloud, the third maximum being compared with the second maximum. Alternatively, the second maximum is preferably transformed by means of the motion vector into a fourth maximum of a fourth surface spanning the fourth point cloud, the fourth maximum being compared with the first maximum.

According to a development of the invention it is provided that the environment is divided into a plurality of cells, the method being carried out on at least one cell of the plurality of cells.

It is thus advantageously possible to carry out the method, in particular the determination of the first point or the first maximum, and the second point assigned to the first point or the second maximum assigned to the first maximum, quickly and reliably.

The plurality of cells is preferably selected such that at least two cells, particularly preferably all cells, have an identical size.

In a particularly preferred embodiment, the method is carried out iteratively for the plurality of cells simultaneously or one after the other. In particular, a similarity value is calculated for each cell of the plurality of cells, the at least one parameter of the odometry device being assessed and / or calibrated using a sum of the similarity values or a mean value of the similarity values.

According to a further development of the invention, it is provided that before the comparison of the first depth recording and the second depth recording with the motion vector, at least one dynamic object is removed from the first depth recording and / or the second depth recording.

A calibration of the odometry device is advantageously better, the more static the surroundings of the motor vehicle are.

The object is also achieved by creating a control device which is created for carrying out a method according to the invention or a method according to one or more of the embodiments described above. In connection with the control device, there are in particular the advantages that have already been explained in connection with the method.

The control device is preferably set up to be operatively connected to the sensor and set up to control it. In addition, the control device is preferably set up to be operatively connected to the odometry device and set up to control it.

The object is also achieved by creating a calibration device which has a sensor for creating depth recordings and a control device according to the invention or a control device according to one or more of the embodiments described above. In connection with the calibration device, there are in particular the advantages that have already been explained in connection with the method and the control device.

The control device is preferably operatively connected to the sensor and is set up to control it.

The object is also achieved by creating a motor vehicle with an odometry device and a calibration device according to the invention or a calibration device according to one or more of the embodiments described above. In connection with the motor vehicle, there are in particular the advantages that have already been explained in connection with the method, the control device and the calibration device.

The control device is preferably operatively connected to the odometry device and is set up to control it.

The motor vehicle is particularly preferably designed as an in particular autonomously driving passenger vehicle. But it is also possible that the motor vehicle is a truck, a commercial vehicle or another motor vehicle.

The invention is explained in more detail below with reference to the drawing.

• 1 eine schematische Darstellung eines Ausführungsbeispiels eines Kraftfahrzeugs,
• 2 ein Flussdiagramm eines Ausführungsbeispiels eines Verfahrens zum Kalibrieren einer Odometrievorrichtung,
• 3 eine schematische Darstellung einer ersten Punktwolke, einer zweiten Punktwolke und einer dritten Punktwolke.

Show:

• 1 a schematic representation of an embodiment of a motor vehicle,
• 2 a flowchart of an embodiment of a method for calibrating an odometry device,
• 3 a schematic representation of a first point cloud, a second point cloud and a third point cloud.

1 shows a schematic representation of an exemplary embodiment of a motor vehicle 1 with a calibration device 3 and an odometry device 5 . The calibration device 3 has a sensor 7th and a control device 9 on.

The sensor 7th is set up to take a depth survey 11th an environment of the motor vehicle 1 to create. Preferably the sensor is 7th set up to take the depth survey 11th the surroundings of the motor vehicle 1 as a point cloud 13th - particularly shown in 3 - to create. The sensor is particularly preferred 7th selected from a group consisting of an SFM vision sensor, a USS sensor, a radar sensor and a lidar sensor.

The odometry device 5 is set up to be a motion vector 15th movement of the motor vehicle 1 to be determined and provided between two points in time. In particular, the odometry device determines 5 based on at least one parameter 17th , selected from a group consisting of a wheel circumference, an abrasion coefficient, a wheelbase, a weather situation and a temperature, in particular an ambient temperature, the motion vector 15th .

The control device 9 is shown here only schematically and is not explicitly shown with the sensor 7th and the odometry device 5 operatively connected and set up for their respective control. The control device 9 is set up in particular to carry out a method for calibrating the odometry device 5 . The procedure is based on 2 explained in more detail.

2 shows a flowchart of an embodiment of the method for calibrating the odometry device 5 of the motor vehicle 1 .

In a first step a), at a first point in time, the sensor 7th a first depth survey 11.1 an environment of the motor vehicle 1 created. The first depth recording preferably has 11.1 a first point cloud 13.1 on.

In a second step b), at a second point in time, the sensor 7th a second depth view 11.2 the surroundings of the motor vehicle 1 created. The second depth recording preferably has 11.2 a second point cloud 13.2 on.

In a third step c) the odometry device 5 the motion vector 15th for movement of the motor vehicle 1 provided between the first time and the second time.

In a fourth step d) the first depth recording is made 11.1 , the second depth view 11.2 and the motion vector 15th compared, being a comparison 19th is created.

In an optional first fourth step d1), at least one dynamic object is preferably obtained from the first depth recording 11.1 and / or the second depth survey 11.2 removed.

Alternatively or additionally, preferably in an optional second fourth step d2), the surroundings are divided into a plurality of cells 21 assigned. The steps following the second, fourth step d2) are performed on at least one cell 21 the majority of cells 21 carried out.

Alternatively or additionally, the first depth recording is preferably carried out in an optional third fourth step d3) 11.1 , especially the first point cloud 13.1 by means of the motion vector 15th transformed, creating a third point cloud 13.3 is obtained.

As an alternative or in addition, the first depth recording is preferably carried out in an optional fourth, fourth step d4) 11.1 , the second depth view 11.2 and the motion vector 15th compared, with a similarity score 23 is determined. In a preferred embodiment, based on the similarity value 23 the comparison 19th created.

The first point cloud is particularly preferred in the optional fourth fourth step d4) 13.1 , the second point cloud 13.2 and the motion vector 15th compared. In particular, the third point cloud 13.3 with the second point cloud 13.2 compared, with preferably at least a third point 25.3 the third point cloud 13.3 , which is in particular by means of the motion vector 15th from the at least one first point 25.1 the first point cloud 13.1 results, in particular by means of an SSIM method, with at least one of the first point 25.1 assigned, second point 25.2 the second point cloud 13.2 is compared.

The at least one first point is particularly preferred 25.1 a first maximum a first the first point cloud 13.1 spanning surface 27.1 certainly. In addition, the at least one first point is preferably used 25.1 associated, second point 25.2 a second maximum, assigned to the at least one first maximum, of a second, the second point cloud 13.2 spanning surface 27.2 certainly. In particular, it is possible to use the first surface 27.1 by means of the motion vector 15th in a third the third point cloud 13.3 transforming spanning surface, the at least one first point 25.1 in the at least one third point 25.3 is transformed.

Alternatively or additionally, the similarity value is preferably used in an optional fifth fourth step d5) 23 with a threshold 29 compared, preferably the comparison 19th based on the similarity value 23 and the threshold 29 is created.

Alternatively or additionally, it is checked in an optional sixth fourth step d6) whether for a predetermined number of cells 21 the majority of cells 21 the first depth survey 11.1 , the second depth view 11.2 and the motion vector 15th were compared.

If the predetermined number of cells 21 the majority of cells 21 have been analyzed, the comparison is preferred 19th created.

If the predetermined number of cells 21 the majority of cells 21 have been analyzed, the optional steps d3) to d5) are carried out again.

In a fifth step e) is based on the comparison 19th at least one parameter 17th the odometry device 5 rated and calibrated.

3 shows a schematic representation of the first point cloud 13.1 , the second point cloud 13.2 and the third point cloud 13.3 .

Identical and functionally identical elements are provided with the same reference symbols in all figures, so that reference is made to the previous descriptions.

In 3 a) is the motor vehicle 1 with a first direction of travel 31.1 shown in a plan view at the first point in time. The surroundings of the motor vehicle 1 is here as the first point cloud 13.1 shown. Furthermore, there is the environment and in particular the first point cloud 13.1 into a plurality of cells 21 , especially a first cell 21.1 and a second cell 21.2 , divided. The first point cloud 13.1 will surface from the first 27.1 overstretched. The first point is particularly preferred 25.1 for calibrating the odometry device 5 analyzed.

In 3 b) is the motor vehicle 1 with a second direction of travel 31.2 shown in a plan view at the second point in time. The surroundings of the motor vehicle 1 is here as the second point cloud 13.2 shown. Furthermore, there is the environment and in particular the second point cloud 13.2 into a plurality of cells 21 , especially a third cell 21.3 and a fourth cell 21.4 , divided. The second point cloud 13.2 is from the second surface 27.2 overstretched. The first point is particularly preferred 25.1 associated, second point 25.2 for calibrating the odometry device 5 analyzed.

the 3 c) and 3 d) each show an overlay of the 3 a) and 3 b) by means of a motion vector 15th . In particular, by means of the motion vector 15th the first point cloud 13.1 into the third point cloud 13.3 transformed, with the first point 25.1 in the third point 25.3 is transformed.

For a clearer representation, the majority of cells have been selected 21 , the first surface 27.1 , the second surface 27.2 and waived the third surface.

In 3 c) are the second point cloud 13.2 and the third point cloud 13.3 identical. From this it can be concluded that that of the odometry device 5 certain motion vector 15th the movement of the motor vehicle 1 correctly represented in the period between the first point in time and the second point in time. The motion vector 15th includes both a translation 33 , as well as a rotation by at least one angle 35 .

In 3 d) are the second point cloud 13.2 - represented by the black filled-in circles - and the third point cloud 13.3 - shown by means of the not filled circles - arranged offset to one another. The second point cloud 13.2 and the third point cloud 13.3 are in particular based on a distance 37 between the second point 25.2 and the third point 25.3 analyzed. From this it can be concluded that that of the odometry device 5 certain motion vector 15th the movement of the motor vehicle 1 not correctly represented in the period between the first point in time and the second point in time.

QUOTES INCLUDED IN THE DESCRIPTION

This list of the documents listed by the applicant was generated automatically and is included solely for the better information of the reader. The list is not part of the German patent or utility model application. The DPMA assumes no liability for any errors or omissions.

Non-patent literature cited

• "Image quality assessment: from error visibility to structural similarity" by Z. Wang et. al., IEEE Transactions on Image Processing, Volume 13, No. 4, April 2004, 600-612 [0014]
• “3D Feature Point Extraction from Lidar Data using a Neural Network” by Y. Feng et. al., The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Volume XLI-B1, 2016 and XXIII ISPRS Congress, 12-19 July 2016 [0020]

Claims (10)

Method for calibrating an odometry device (5) of a motor vehicle (1), wherein - At a first point in time, a first depth recording (11.1) of the surroundings of the motor vehicle (1) is created by means of a sensor (7), wherein - At a second point in time, a second depth recording (11.2) of the surroundings of the motor vehicle (1) is created by means of the sensor (7), wherein - A motion vector (15) for a movement of the motor vehicle (1) between the first point in time and the second point in time is provided by the odometry device (5), wherein - The first depth recording (11.1), the second depth recording (11.2) and the motion vector (15) are compared, with - Based on the comparison (19) at least one parameter (17) of the odometry device (5) is evaluated and calibrated. Procedure according to Claim 1 , the first depth recording (11.1) having a first point cloud (13.1), the second depth recording (11.2) having a second point cloud (13.2). The method according to any one of the preceding claims, wherein - A similarity value (23) is determined for the first depth recording (11.1), the second depth recording (11.2) and the motion vector (15), wherein - The similarity value (23) is compared with a threshold value (29), wherein - Based on the comparison (19) of the similarity value (23) with the threshold value (29), the at least one parameter (17) of the odometry device (5) is evaluated and calibrated. The method according to any one of the preceding claims, wherein - At least one first point (25.1) of the first point cloud (13.1) is transformed by means of the motion vector (15), whereby at least one third point (25.3) is obtained, wherein - The at least one third point (25.3) is compared with a second point (25.2) of the second point cloud (13.2) assigned to the at least one first point (25.1), wherein - Based on the comparison (19), the at least one parameter (17) of the odometry device (5) is evaluated and calibrated. The method according to any one of the preceding claims, wherein - A first maximum of a first surface (27.1) spanning the first point cloud (13.1) is determined as the at least one first point (25.1), wherein - as the second point (25.2) assigned to the at least one first point (25.1), a second maximum, assigned to the at least one first maximum, of a second surface (27.2) spanning the second point cloud (13.2) is determined. Method according to one of the preceding claims, wherein the environment is divided into a plurality of cells (21), the method being carried out on at least one cell (21) of the plurality of cells (21). Method according to one of the preceding claims, wherein before the comparison (19) of the first depth recording (11.1) and the second depth recording (11.2) with the motion vector (15) at least one dynamic object from the first depth recording (11.1) and / or the second depth recording (11.2) is removed. Control device (9), set up to carry out a method according to one of the preceding claims. Calibration device (3) with a sensor (7) for making depth recordings (11) and a control device (9) according to Claim 8 . Motor vehicle (1) with an odometry device (5) and a calibration device (3) according to Claim 9 .

Images