Method for estimating a correction angle of a radar sensor of a motor vehicle

Abstract

A method for estimating a correction angle (γ1, γ4) of a radar sensor (12) of a motor vehicle, in which the correction angle taking into account the misalignment of the radar sensor is calculated by statistically analyzing positioning data received by the radar sensor (12), characterized in that the positioning angle range (W) of the radar sensor (12) is divided into a plurality of sectors (S1-S4), and the positioning data is statistically analyzed separately for the different sectors, thereby being able to obtain its own correction angle (γ1, γ4) for each sector.

Description

Technical Field

[0001] The invention relates to a method for estimating a correction angle of a radar sensor of a motor vehicle, in which method a correction angle taking into account misalignment of the radar sensor is calculated by statistically evaluating positioning data received by the radar sensor. Background Art

[0002] In motor vehicles, angle-resolving radar sensors are used for various assistance functions, such as automated distance control, collision warning and emergency braking systems, and in the future also for fully autonomous driving.

[0003] In most assistance systems of this type, at least one radar sensor is mounted at the front of the vehicle, whose optical axis coincides with the longitudinal axis of the vehicle, so that the positioning angle measured by the radar sensor for each object indicates the angular extension of the relevant object relative to the longitudinal axis of the host vehicle.

[0004] Since some assistance functions require both the azimuth angle and the elevation angle of the located object, the radar sensor should also be aligned in such a way that its optical axis extends horizontally and can serve as a reference for angle measurement in elevation.

[0005] If the optical axis of the radar sensor is not correctly aligned, for example due to manufacturing tolerances of the radar sensor, installation errors when installing the radar sensor in the vehicle, or also due to mechanical effects such as parking bumps during vehicle operation, the results of all angle measurements are distorted by the misalignment angle of the radar sensor.

[0006] To avoid erroneous angle measurements and correspondingly inaccurate estimates of the traffic situation, radar sensors can be calibrated after installation in the vehicle using relatively complex and time-consuming measurement procedures so that any existing alignment errors can be measured and subsequently corrected computationally during data evaluation.

[0007] Methods are also known which enable adjustment errors to be controlled and, if necessary, corrected even during continued operation of the vehicle. An example of such a method is described in DE 10 2006 045 165 A1.

[0008] These methods generally involve statistical evaluation of the positioning data of the objects located by the radar sensor. For example, if the azimuth of a vehicle traveling immediately behind in the host vehicle's lane is measured, then the temporal average value of this azimuth should converge to 0° if the sensor is correctly aligned, since the vehicle traveling immediately ahead is offset slightly to the right or slightly to the left relative to the host vehicle with equal probability.

[0009] Another method for measuring alignment errors is based on the following consideration: stationary objects at the edge of the road do not change their lateral relative position during driving, at least when the host vehicle does not make any lateral relative movements. If the azimuth of such an object is tracked for a certain period of time during driving, the azimuth has a characteristic time correlation due to the translation that occurs during driving. If there is an alignment error, this time correlation changes and an apparent movement of the actually stationary object in the lateral direction of the vehicle is detected. Based on this effect, the misalignment can be determined quantitatively. However, this result may be distorted by measurement inaccuracies. Even if the host vehicle makes a slight lateral movement during the measurement time, it may lead to measurement errors as long as the lateral movement of the host vehicle is not compensated in the calculation. Therefore, in this method, in order to increase the accuracy, data from multiple objects are usually received and the alignment errors are then compensated by static averaging.

[0010] As the complexity of the assistance functions increases, the accuracy requirements for the angle measurement also increase. In particular, when the radar sensor is oriented forward in the direction of travel in the sense of the "forward-looking" behavior of the assistance function, it is desirable to also be able to observe traffic situations at a relatively large distance in front of the host vehicle. However, since the lateral position of a located object in a Cartesian coordinate system is proportional to the azimuth angle and the distance to the object, the greater the distance to the object, the greater the error in the azimuth angle measurement. Summary of the invention

[0011] The object of the present invention is therefore to specify a method of the type mentioned in the introduction which allows the correction angle to be estimated with greater accuracy.

[0012] According to the invention, this object is achieved by dividing the positioning angle range of the radar sensor into a plurality of sectors and statistically evaluating the positioning data of the different sectors individually, thereby obtaining a correction angle for each sector.

[0013] The method according to the invention takes into account the fact that errors in the angle measurement can be caused not only by misalignment of the sensor, but also by deviations in the beam path of the radar beam and / or by systematic angle errors caused by errors in the evaluation of the measurement signal. Such systematic angle errors occur, for example, when the radar sensor has a focusing lens with certain manufacturing tolerances or is installed, for example, behind the bumper of the vehicle, and the shape of the bumper or, if necessary, dust causes a deflection of the radar beam similar to that in a prism. Other possible causes for systematic angle errors are propagation time errors in radar signals, which are transmitted and received, for example, by different antenna elements of the radar sensor in the case of digital beam forming.

[0014] Unlike errors caused by radar sensor misalignment, systematic angle errors are usually angle-dependent. That is, the degree to which the angle measurement is distorted is related to the corresponding positioning angle of the located object.

[0015] However, in conventional methods for measuring alignment errors, the data of the object distributed over the entire positioning angle range are averaged, so that angle-dependent systematic errors disappear in the statistical evaluation and are not correctly reflected by a correction angle which essentially corresponds to the sensor misalignment angle.

[0016] In contrast, in the method according to the invention, by dividing the locating angle range into a plurality of sectors and by separate, sector-by-sector evaluation, it is achieved that the correction angles obtained for the different sectors reflect not only the misalignment angles but also the systematic errors for the angles located in the relevant sector. Although statistical averaging is still performed here, it is only performed over individual sectors, so that the result describes the systematic errors that actually occur in the sensor for this object with greater accuracy.

[0017] If the position of the located object is subsequently to be determined within the scope of an assistance function, the correction angle belonging to the sector in which the object is located is used for the angle correction.

[0018] In this way, errors in the angle measurement can be compensated with greater accuracy.

[0019] This method can be used not only for angle measurement in azimuth but also for angle measurement in elevation.

[0020] Advantageous embodiments and developments of the invention are specified in the dependent claims.

[0021] Since the function which describes the system angle error as a function of the angle is generally continuous, the method can be improved by interpolating between the correction angles for two adjacent sectors if the located object is not located exactly on the angular bisector of the sector.

[0022] The greater the number of sectors and the smaller the angular range occupied by each sector, the more accurately the correction angle determined for these sectors agrees with the actual angular error. On the other hand, however, as the size of the sector decreases, the probability of locating the object in the associated sensor also decreases, so that more time is required to collect enough data for statistical analysis.

[0023] The method according to the invention also allows, within certain restrictions, the quantitative determination of the systematic angular error as a function of the angle. At least the variation of the angular error relative to a fixed reference value can be measured. If, however, the angular error contains constant components that are the same for all located objects, regardless of the azimuth angle, these components have the same effect on the distortion of the angle measurement as alignment errors of the sensor, so that these two sources of error cannot be distinguished. However, if the deviations of the angular error from an arbitrary fixed reference value are taken into account, it is possible to compare the angle-dependent systematic errors in different sectors.

[0024] It can be determined for two given, not necessarily adjacent sectors or also for a group of three or more sectors whether the systematic angular errors in these sectors are identical. If this is the case, the sectors can be combined into a larger angular range, which can then be statistically evaluated. This increases the probability of locating an object within the larger angular range and shortens the measurement duration required to receive enough samples of locating data for a convincing statistical analysis. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] The embodiments are further described below with reference to the accompanying drawings.

[0026] The accompanying drawings show:

[0027] Figure 1 : A plan view of a motor vehicle having a radar sensor with an alignment error, in combination with a diagram of the positioning angle range of the radar sensor;

[0028] Figure 2 : Block diagram of a radar sensor, with which the method according to the invention can be implemented;

[0029] Figure 3 and Figure 4 : A diagram for illustrating a method for measuring an alignment error of a radar sensor; and

[0030] Figure 5 : Flow chart of an embodiment of the method according to the present invention. DETAILED DESCRIPTION

[0031] exist Figure 1 1 shows a plan view of a motor vehicle 10 having a radar sensor 12 pointing forward in the direction of travel.

[0032] Axis A describes the longitudinal center axis extending in the direction of travel of motor vehicle 10. In the example shown, radar sensor 12 is not correctly aligned with axis A but has a certain alignment error, i.e., its optical axis B forms an angle δ with axis A, which describes its alignment error.

[0033] Radar sensor 12 has an angle range W, which is illustrated here as a circular sector symmetrical to optical axis B.

[0034] Furthermore, a Cartesian coordinate system is plotted with an axis x oriented in the direction of travel and an axis y oriented in the transverse direction of the vehicle. Point P indicates the actual position of an object located by radar sensor 12 in this Cartesian coordinate system.

[0035] An axis AP connects the radar sensor 12 with the point P. The angle between the axis A and AP is the true azimuth of the object at the point P.

[0036] However, due to the alignment error, the radar sensor 12 “sees” the object at a point P′ on an axis AP′ which is twisted by an angle δ relative to the axis AP.

[0037] Point Q on axis AQ describes the actual position of another object located by radar sensor 12. Due to alignment errors, this object is also seen by radar sensor 12 at a fictitious point Q' on axis AQ' which is rotated by angle δ relative to axis AQ.

[0038] In the example shown, it is assumed that radar sensor 12 also has a systematic angular error which results in an object that is actually at point P being positioned at position P″ on axis AP″.

[0039] In the case where the object is actually at point Q, this angular error has the effect of positioning the object at point Q" on axis AQ".

[0040] Unlike the alignment error described by the angle δ, which is the same for all located objects, regardless of the azimuth angle at which they are located, the systematic angle error is angle-dependent. Figure 1 In , the angular error is specified for an object at point P by the angle between the axes AP' and AP", while the system angular error is specified for an object at point Q by the angle between the axes AQ' and AQ". It can be seen that these angular errors are different from one another.

[0041] As will be explained in detail below, methods are known with which a correction angle can be calculated which is equal to the angle δ and which can correct the alignment error, but under the assumption that there is no angle-dependent systematic angular error. Figure 1If this correction is carried out in the case of a radar sensor 12 with a fixed position, a correction angle that does not exactly correspond to the angle δ will be obtained. If an average is taken for a large number of objects that are substantially evenly distributed over the positioning angle range W, a correction angle is obtained that is composed of the angle δ (corresponding to the alignment error) and the average value of the system angle error. Even if the average value of the system angle error is assumed to be 0°, only the differences between the points P and P' and between Q and Q' can be eliminated when this correction is carried out, but the deviations between P' and P" and between Q' and Q" due to the angle-dependent system angle error remain unchanged.

[0042] To improve the correction accuracy, Figure 1 The positioning angle range W is divided into a plurality of (four in the example shown) sectors S1, S2, S3 and S4, in which all of these sectors have the same size. The alignment error is determined separately for each of the sectors. That is, in order to determine the alignment error for one of the sectors by statistically analyzing the positioning data, only the data of the objects located in the sector are considered. In this way, a virtual optical axis is obtained for each sector, which is twisted by a correction angle relative to the axis A. Figure 1 , virtual axes B1 and B4 and the associated correction angles γ1 and γ4 are shown for sectors S1 and S4. Axes B1 and B4 are virtual optical axes in the sense that they do not describe the actual misalignment of radar sensor 12, but rather a false misalignment taking into account the angular errors applicable to the relevant sectors.

[0043] If in this way, after a certain measuring time, during which sufficient statistical data are collected for each sector, a correction angle γ1-γ4 is determined for each sector and the data is provided to the radar sensor 12 for auxiliary functions, the azimuth angle is corrected for each located object by a correction angle, which applies to the sector S1-S4 in which the object is located.

[0044] In the example shown, the axis AQ" which describes the positioning angle of the object at position Q" is approximately in the middle of sector S4. In this case, the associated correction angle γ4 will be used directly to correct the alignment error and the angular error. In contrast, in the case of an object at position P, the axis AP" is located closer to the edge of sector S1. Based on the reasonable assumption that the angular error and thus the correction angle change continuously over the entire positioning angle range W, in this case the correction angle γ1 is not used directly, but rather a correction angle obtained by interpolation between γ1 and the corresponding correction angle for sector S2.

[0045] exist Figure 2, the essential components of a radar sensor 12 with which the above-described method can be implemented are shown in the form of a block diagram. A transmitting and receiving unit 16 with a subordinate pre-evaluation unit provides, in each measuring cycle, for each located object a distance d, a relative speed of the object (which is irrelevant here and therefore not shown), and an azimuth angle α, which may be distorted due to alignment errors and angular errors. The distance d and the azimuth angle α describe the assumed position of the located object in polar coordinates. A conversion unit 18 converts the polar coordinates into Cartesian coordinates and provides, inter alia, a coordinate y for each object, which describes the lateral position of the object relative to the axis A.

[0046] In the estimation module 20 , a correction angle γ is determined for each sector S1 - S4 of the positioning angle range. In this case, only the positioning data of the object located in the sector for which the correction angle is determined are taken into account.

[0047] The correction module 22 then corrects the measured azimuth angle α by the corresponding correction angle γ, using interpolation between the two correction angles if necessary, and provides the corrected rectangular coordinate x * y * , the coordinates describe the real position P or Q of each object with higher precision.

[0048] The functions of the above-mentioned components of the radar sensor are controlled by a control unit 24 .

[0049] Now according to Figure 3 and Figure 4 An example of a method that can be used to estimate the correction angle γ is shown.

[0050] exist Figure 3 1 shows a traffic situation in which a vehicle 10 equipped with a radar sensor 12 follows another vehicle 26 that is traveling directly ahead in its own lane. Radar sensor 12 has an alignment error that is described by angle δ. If, as in the example shown, the actual position of vehicle 26 is exactly on axis A, radar sensor 12 measures an azimuth angle that corresponds to angle δ.

[0051] Figure 4The situation is shown at a later point in time. Vehicle 26 is no longer keeping the exact course and is now slightly to the left of axis A. If vehicle 26 is tracked over a longer period of time and the azimuth is measured repeatedly in each measuring cycle, the measured azimuths are evenly distributed around the value δ, i.e. vehicle 26 randomly deviates sometimes slightly to the right of axis A and then slightly to the left of axis A. If a smoothed average is calculated from the measured azimuths, this average converges over time to a limit value which describes the sought correction angle with high accuracy, which consists of angle δ and, if necessary, a systematic angle error with respect to the 0° direction of the radar sensor. The longer the measuring time and the correspondingly larger the statistical basis, the more accurately the limit value obtained reflects the "correct" correction angle.

[0052] For the sectors farther from the axis A in the positioning angle range W, for example Figure 1 Sectors S1 and S4 in the example are suitable for other methods, which also apply the same Figure 3 and Figure 4 To elaborate.

[0053] exist Figure 3 In one of the sectors, the radar sensor 12 locates a stationary object 28 which is at the edge of the road and is Figure 3 There is still a relatively large distance between the vehicle 10 and the host vehicle 10. Figure 4 , vehicles 10 and 26 have continued to move such that the distance to object 28 (in the X direction) has been greatly reduced.

[0054] Conversion unit 18 ( Figure 2 ) mainly calculates the coordinates of object 28. If host vehicle 10 remains constant on axis A, coordinate y should not change over time, because object 28 is stationary (stationary objects are identified in that their radial speed relative to the sensor multiplied by the cosine angle is equal to and opposite to the speed of travel of the host vehicle). However, due to alignment errors, conversion module 18 does not calculate the true y coordinate of object 28, but a false y coordinate relative to optical axis B of the radar sensor. Figure 4 The false y coordinates in Figure 3 If the increase in the false y coordinate is plotted against the distance travelled by the host vehicle 10, a straight line is obtained (after correction for the host vehicle's own movement, if necessary) whose slope is exactly equal to the angle δ or the mechanical alignment error plus the angular error for the relevant sector. In this way, an estimate for the correction angle can already be obtained based on a single stationary object 28. However, this estimate still has static measurement errors. By statistically evaluating a plurality of successively located stationary objects, the statistical errors can be reduced and the accuracy can be improved.

[0055] In addition, Figure 3 and 4 As can be seen in , the azimuth angle of object 28 measured by the radar sensor increases as vehicle 10 approaches the object further. In practice, therefore, the measured azimuth angle sweeps through a plurality of sectors (S2, S1) of the orientation angle range, so that separate evaluations can be performed for different sectors based on a single object, and a plurality of correction angles are obtained accordingly, which differ from one another due to different angular errors in the relevant sectors.

[0056] Now according to Figure 5 A possible method sequence in the method according to the invention is explained with reference to the flowchart shown in . The method is implemented, for example, in control unit 24 of a radar sensor.

[0057] In step ST1, the positioning angle range W is divided into sectors, for example, according to Figure 1 It is divided into four sectors S1-S4, or can be selectively divided into a larger number of sectors.

[0058] In step ST2 , it is checked whether the systematic angle error of radar sensor 12 for the sector determined in step ST1 is known. If this is the case, in step ST3 , the individual sectors having the same angle error are combined into a single (possibly discontinuous) sector.

[0059] If the angle error is still unknown (N in step ST2 ), step ST3 is skipped.

[0060] Then, in step ST4, the positioning data of the located object is recorded individually on a sector-by-sector basis.

[0061] In step ST5, check whether the number of objects recorded in step ST4 in each sector to locate data has reached a certain minimum value, so that the sample for statistical analysis has sufficient quantity. As long as this is not the case (N), jump back to step ST4 and continue data recording.

[0062] If all sectors have reached a sufficient sample size, then in step ST6, for example, Figure 3 and Figure 4 One of the methods described estimates the alignment error on a sector by sector basis.

[0063] The correction angles obtained in step ST6 are then averaged in step ST7, weighted according to the number of samples in the individual sectors. This results in an average correction angle being determined for the entire positioning angle range W. This average correction angle contains, on the one hand, mechanical alignment errors of radar sensor 12 and, on the other hand, a constant, angle-independent component of the system angle error.

[0064] In step ST8 , the angle error is then calculated for each individual sector by subtracting the correction angle obtained in step ST6 from the average correction angle obtained in step ST7 .

[0065] In step ST9, the correction angle obtained in step ST6 is compared with the correction angles for the same sector that were stored in advance, and a check is performed to see whether the correction angles are stable in all sectors, i.e., whether the deviations between the correction angles obtained in the most recent time for the same sector are within a predefined tolerance range. If this is not the case, all sectors with the same angular error are merged again in step ST10. This step is a repetition of step ST3, but now taking into account the angular error that was only obtained in step ST8 or updated if necessary.

[0066] The minimum sample size for each sector is increased in step ST11, and the positioning data for each sector is re-recorded in step ST12.

[0067] In step ST13, a check is made whether a (larger) minimum sample size or convergence has been reached. If this is not yet the case (N), in step ST12, the recording of positioning data continues, and steps ST12 and ST13 are repeated until the minimum sample size is reached. If this is the case, a jump is made back to step ST6 and the loop with steps ST6 to ST13 is repeated periodically until it is determined in step ST9 that the correction angles are stable in all sectors. It will be appreciated that in this step ST9, the correction angles obtained in the different repetitions of the loop from ST6 to ST13 are compared with one another. If the sequence of correction angles is sufficiently stable (Y), the method ends with step ST14.

[0068] By merging sectors in steps ST3 and ST10 , it is achieved that more objects can be located in each sector within a given time interval, so that the method converges more quickly or statistical fluctuations are further suppressed in sectors enlarged in this way.

[0069] Optionally, in step ST9, if the correction angle is not yet stable in all sectors, it can be additionally checked whether the correction angles of at least two or more sectors show a certain degree of convergence. If this is not the case, step ST10 is skipped and only further data is collected in the loop ST11-ST6-ST9. The merging of sectors with the same angular error is then performed only for those sectors whose angular errors have a sufficient degree of stability and reliability.

[0070] The method described here can be repeated at certain time intervals during the service life of motor vehicle 10 in order to control the calibration of the radar sensor with respect to alignment errors and system angle errors. The method can also be performed continuously in the background while the radar sensor provides data for the assistance function.

[0071] The division of the positioning angle range into sectors in step ST1 can also be varied in different repetitions of the method. For example, it is possible to start with a relatively small number of sectors in order to obtain convincing static results as quickly as possible, even if the detection of the angle-dependent angular errors is relatively rough, and then to work with a larger number of sectors in a second step in order to determine a curve that describes the angular dependence of the systematic error with a higher resolution. If positioning data are stored in different program runs, the data base can be increased in a program run with a larger number of sectors by revisiting positioning data from a previous program run with a smaller number of sectors by subsequently dividing into new sectors.

Claims

1. A method for estimating a correction angle (γ1, γ4) in a radar sensor (12) of a motor vehicle (10), wherein: The method comprises: calculating a correction angle by statistically analyzing the positioning data received by the radar sensor (12), wherein the correction angle takes into account the misalignment of the radar sensor, and is characterized in that the positioning angle range (W) of the radar sensor (12) is divided into a plurality of sectors (S1-S4), and the statistical analysis is performed on the positioning data separately for these different sectors, thereby obtaining a correction angle (γ1, γ4) for each sector. Therein, an angle-dependent component of a system angle error of the radar sensor (12) is stored for each sector (S1-S4), and wherein, if the stored angle-dependent component is identical for two or more sectors, the division of the sectors is changed in such a way that sectors with identical values ​​are merged into one larger sector.

2. The method according to claim 1, wherein: A correction angle obtained by interpolating between the correction angles obtained for the respective sectors (S1-S4) is used to correct the positioning angle data of the object.

3. The method according to claim 1 or 2, wherein: The angle-dependent component of the system angular error is determined as a function of the correction angle (γi) calculated during the method.

4. The method according to claim 1 or 2, wherein: The merging of sectors is performed only for those sectors whose angular errors have a sufficient degree of stability and reliability.

5. The method according to claim 1 or 2, wherein: The method is repeated and the division of the positioning angle range into sectors is varied in different repetitions of the method.

Images