DE10200944A1 - Automatic vehicle distance sensor output correction involves deriving correction angle from phase difference between reflection signals from two objects enclosing angle with vehicle - Google Patents

Abstract

The method involves emitting an electromagnetic pulse(s), detecting a reflected signal amplitude and a transverse distance of the reflecting object and determining the correction value by comparing transverse distance values. Two reflection signals are simultaneously detected from two objects enclosing an angle with the vehicle, two transverse distance values are determined and the correction angle derived from the signals' phase difference. The method involves emitting at least one electromagnetic pulse, detecting the amplitude of a reflected signal depending on an angle relative to a distance sensor axis, determining a transverse distance of the reflecting object from the vehicle's path depending on the transition time and amplitude of the signal and the angle and determining the correction value by comparing several transverse distance values. First and second reflection signals are simultaneously detected from first and second objects enclosing an angle with the vehicle and first and second transverse distance values are determined and determining the correction angle from a phase difference between the signals.

Description

The invention relates to a method for the automatic correction of Output values of a distance sensor in a vehicle according to the Preamble of claim 1.

For the correct functioning of driver assistance systems based on sensors, the information about the surrounding traffic is the adjustment of this Crucial sensors. An example of such a system is the ACC cruise control. In general, it is required that the Sensor axis is parallel to the direction of travel. Such a sensor is considered to be adjusted designated. If the system automatically detects a sensor misalignment, see above can correct the positions of the objects detected by the sensor afterwards system, or if a critical limit is exceeded switch off automatically.

A known measuring principle for misalignment detection is based on the fact that with an ideally adjusted radar sensor, the measured transverse distances of a stationary target have the same value for all distances to this target (apart from the inevitable measurement noise). However, the condition for this is that the sensor vehicle moves on a straight line. On the other hand, if a vehicle with a misaligned sensor approaches a stationary target in a straight line, the measured transverse distances initially deviate from the ideal value q ₀ = y _Z - y ₀ , which an adjusted sensor would measure. As the rapprochement approaches, these deviations become smaller and smaller. The actual transverse distance is therefore extrapolated by means of linear regression from a sequence of continuously calculated transverse distance values.

A disadvantage of this prior art has been that the Limits of the method of linear regression are reached when that Sensor vehicle deviates from the linear movement. A non-linear Movement of the vehicle means that even with an ideally adjusted sensor the measured transverse distance of a standing target as a function of the measuring distance fluctuates because the horizontal line of sight of the radar sensor changes over time changed. Using the linear regression method described above, the Correction angle can no longer be determined.

The object of the present invention is to provide a method with which Output values of a distance sensor of a motor vehicle also then are automatically corrected if the vehicle does not move in a straight line.

This object is achieved by a method for automatic correction of output values of a distance sensor at a Vehicle according to claim 1. Preferred embodiments of the invention are Subject of the subclaims.

The invention is based on the idea that the phase difference of reflection signals, which are thrown back by two objects. The Phase difference depends u. a. on the distance between the distance sensor and the objects.

The inventive method for the automatic correction of Output values of a distance sensor in a vehicle that is on a Moved path, which comprises the steps: sending at least one electromagnetic pulse by the distance sensor, detecting an amplitude an electromagnetic reflection signal reflected from an object, that was generated by the at least one electromagnetic pulse, in Dependence on an angle with respect to a distance sensor axis, determining a transverse distance of the object from the path of the vehicle in Dependence on the transit time and the amplitude of the reflection signal and the Angle, comparing several cross-distance values and determining one Correction angle for the distance sensor from the comparison result is thereby characterized in that at the same time a first reflection signal from a first Object and a second reflection signal from a second object is detected and a first transverse distance value and a second transverse distance value is determined, the first object and the second object with the Vehicle include an angle, a phase difference between the first and the second reflection signal is determined and the correction angle for the Distance sensor is determined from the phase difference.

In particular, the path of the vehicle becomes retrospective from the course over time the phase difference is calculated. This makes it possible to α estimate the deviations of the vehicle movement from the ideal Reflect straight lines. With the estimated vehicle helix angle, the measured transverse distance of the two standing targets are corrected so that it is in has a shape that corresponds to a straight-line vehicle. This Values can then again be obtained using state-of-the-art methods evaluate.

An advantage of the invention is that the detection of misalignment is robust against fluctuations in vehicle movement, so that even the Evaluation of a single target trajectory an exact correction angle supplies. Applications that were previously not possible can be realized with it, namely rapid detection of misalignment, e.g. B. after a parking lot, and the Quality inspection of the systems for sensor adjustment in production on one Reference section.

Further features and advantages of the invention will appear from the following Description of preferred embodiments, reference being made will refer to the attached drawings.

Fig. 1 shows schematically a vehicle with a misaligned distance sensor on a straight line.

Fig. 2 is used to explain the evaluation of a sequence obtained with a misaligned distance sensor distance values according to the prior art.

Fig. 3 shows schematically a vehicle with poorly adjusted distance sensor on a curved path.

Fig. 4 shows the evaluation of measurement signals a misaligned distance sensor according to the prior art.

Fig. 5 shows the determination of the present invention shows a vehicle slip angle from measurement signals of a distance sensor.

Fig. 6 shows the conversion of measurement signals according to the invention a curved path to a linear path followed by the determination of a regression line with a known distance between the distance sensor and the subject.

FIG. 7 shows the determination of a vehicle helix angle according to the invention from measurement signals of a distance sensor.

In Fig. 1, a vehicle 1 is shown, which comprises a distance sensor. At least one electromagnetic pulse is emitted by the distance sensor. The amplitude of an electromagnetic reflection signal reflected by an object 3 , which was generated by the at least one electromagnetic pulse, is recorded as a function of an angle with respect to a distance sensor axis 4 .

The distance sensor can be a rotating radar antenna, over which a short one Radar pulse is transmitted and received. Depending on the angular position of the rotating antenna in the lane level, the direction can be determined from which reflects the signal the most. From the transit time of the reflection signal this results in the distance of the object from the vehicle. For this Total distance and the angle at which the maximum amplitude was detected, can also be a transverse distance q of the object from the web of the vehicle.

The same measurement method can also be used with other radar devices carry out. For example, instead of one rotating radar antenna, several, z. B. 3 fixed antennas are used, their respective Main lobe axes form an angle with each other in pairs. The The measuring principle in this case is the same as above.

It is crucial for the calculation of the transverse distance of the object 3 that the distance sensor axis 4 coincides with the axis of symmetry 2 of the vehicle 1 . In order to be able to recognize possible misalignment angles δ _r between the distance sensor axis 4 and the axis of symmetry 2 of the vehicle 1 and to be able to correct the calculated transverse distance values, transverse distance values at several driving positions are calculated and temporarily stored.

The misalignment angle δ _{r is} determined from the comparison of the sequence of transverse distance values. In FIG. 2 it is shown how from the sequence of individual cross-distance values, a correction angle δ 5 _r is determined for the distance sensor. The determined transverse distance is plotted on the ordinate axis (e.g. in meters), the time of measurement and calculation is plotted on the abscissa (e.g. in seconds).

The transverse distance values 5 are plotted over time as measuring points.

A linear regression is then carried out, ie a regression line 6 is laid through the measurement points.

The measured transverse distance q _mess is composed of the ideal value q ₀ and a second component q _deju , which is caused by the correction _angle δ:
q _{deju ≍} δ.Target distance (valid for small angles), so in total:
q _mess = q ₀ + δ. target distance + measurement noise.

The measured transverse distance q _mess thus depends linearly on the target distance, the correction angle δ determining the slope. The following convention applies to the sign for δ:
δ <0 °: misalignment to the left; δ> 0 °: misalignment to the right.

A straight line is drawn using N measuring points, which are composed of the measured distances to the target d _mess (i) and the measured transverse distances of the target q _mess (i) (i = 1,..., N). The slope of this regression line corresponds to the tangent of the estimated correction angle δ _r . For the special case of an ideally adjusted sensor, a horizontal compensation line would be obtained, which corresponds to a correction angle of zero degrees. The ordinate section of the straight line corresponds to the transverse distance q _{0 of} the standing target, which an ideally adjusted sensor would measure.

When calculating the regression line, it is assumed that the angular noise of the sensor is independent of the distance from the target, the noise of the measured transverse distance is therefore proportional to this distance (valid for small angles). From the intermediate sizes
S = Σ (1 / (d _mess (i)) ² )
S _X = Σ (1 / d _mess (i))
S _Y = Σ (q _mess (i) / (d _mess (i)) ² )
S _XY = Σ (q _mess (i) / d _mess (i))
S _XX = N
can use the parameters of the regression line
δ _r = arctan [(SS _XY - S .S _{X Y)} / (SS _XX - S _X. S _X)]
q _{0_r} = (S _XX .S _Y - S _X .S _XY ) / (SS _XX - S _X .S _X ),
can be calculated, where δ _{r is} the angle of the slope of the straight line and q _{0_r is} the transverse distance determined.

As a result, the misalignment angle δ _{r of} the distance sensor axis 4 is obtained as desired. This misalignment angle δ _r flows into the subsequent processing of the output signals of the distance sensor as a correction angle δ _r a. On the other hand, the representation in FIG. 1 gives the correct transverse distance q0 of the object 3 from the path on which the vehicle is traveling. The correct transverse distance q0 is the intersection of the regression line with the Auer distance axis at time zero. In Fig. 1, both the correct transverse distance value q0 is shown as well as the measured false transverse distance q.

In Fig. 3, a vehicle is shown with a misaligned distance sensor on a curved track 7. The instantaneous angle between the curved path 7 and its temporal average 7 a is denoted by α. The vehicle travels on a roll course in FIG. 3.

In order to allow a safe distance measurement and to be able to correct a possible misalignment angle δ between the vehicle axis 2 and the distance sensor axis 4 independently of the momentary movement of the vehicle on the curved path 7 , two reflection signals from one object each are evaluated according to the invention. The two pulses from the distance sensor for generating reflection signals of an object 3 a and an object 3 b are designated 8 and 9 in FIG. 3.

As in the prior art described above, the reflection signal of an object is also detected in the method according to the invention. In contrast to the prior art, however, a first reflection signal from a first object 3 a and a second reflection signal from a second object 3 b are detected simultaneously according to the invention. It is assumed in the following that the first object 3 a and the second object 3 b enclose the vehicle with an angle, since otherwise the reflection signals could not be distinguished. As usual, a first transverse distance value and a second transverse distance value are determined from the two reflection signals.

In addition, however, a phase difference between the first and the second reflection signal is also determined. This phase difference between the first and the second reflection signal contains information about the distance of the distance sensor from the two objects 3 a and 3 b. Therefore, the correction angle δ _r for the distance sensor can be determined from the phase difference.

The phase difference between the two reflection signals can also be used to subsequently calculate the path 7 of the vehicle.

The details of the method are explained below with reference to FIGS. 4 to 7.

In the prior art, even minimal deviations from the straight-line movement lead to gross incorrect estimates of the correction angle. Fig. 4 shows the result of a simulation calculation, it was assumed cm in which a fluctuation amplitude of the vehicle 20. In addition, this movement was overlaid on a curve with a radius R = 50 km. A Gaussian noise for distance and angle was assumed for the simulated measurement data with the standard deviations σ _x = 1.0 m and σ _φ = 0.3 °. A correction angle of + 0.3 ° was also specified.

If the method of linear regression is applied to the measurement data 10 , an estimate of the correction angle of 0.74 ° is obtained from the regression line 11 , which represents a considerable deviation from the actual value. In order to arrive at better estimates, the individual results of a large number of target trajectories are averaged, but this generally requires very long measurement times of up to a few hours before a final result is available. This is usually still very buggy.

With the method according to the invention one of the vehicle movement can be perform an independent estimation of the correction angle. requirement is, that two stationary targets are detected by the sensor at the same time, but they are as separate goals must exist.

The basic idea of the phase correction method is to use the Reflection signals of the two targets to extract the phase difference, which u. a. depends on the position of the sensor to the targets. From the course of this The phase difference can then be calculated back to the vehicle movement. In particular, the vehicle helix angle α can be estimated, in which the deviations of the vehicle movement from the ideal straight line reflect. With the estimated vehicle helix angle, the measured Corrected the cross distance of the two stand targets so that it is in a form that corresponds to a straight-line vehicle. So that's the simple linear Regression applicable again.

The following relationship in the baseband applies to the complex-value received signal S _{1 of} a reflector with the complex-value backscatter amplitude R ₁ (at a distance d ₁ from the sensor under the aspect angle φ1), the influence of the antenna diagram A (φ ₁ ) on amplitude and phase also being taken into account here is. The weakening of the amplitude due to the transit time is generally described by g (d ₁ ), which is, however, not essential in the following. S ₀ is the amplitude of the transmission signal, λ the wavelength of the carrier signal.
S ₁ (d ₁ , φ ₁ ) = S ₀ .R ₁ (φ ₁ ) .g (d ₁ ) .exp (j.4π.d ₁ /λ).A(φ ₁ )

The same applies to the second stand goal:
S ₂ (d ₂ , φ ₂ ) = S ₀ .R2 (φ ₂ ) .g (d ₂ ) .exp (j.4π.d ₂ /λ).A(φ ₂ )

Since the changes in the distances d ₁ and d ₂ are very large in comparison to the wavelength ia, the phases of S ₁ and S ₂ fluctuate very strongly, so that the phase difference of S ₁ and S ₂ lends itself to further evaluation from the outset.

For the special case that both spreaders sit at the same longitudinal coordinate x _T and that the sensor on the center perpendicular of the connecting line of the spreaders approaches (ie straight!), The distances d ₁ and d _{2 are} always the same. The two spreaders and the sensor form an isosceles triangle at all times. Only when the sensor deviates from this ideal movement does a difference of d ₁ and d ₂ occur, which is reflected in the phase difference. This clearly shows that there is information about the sensor movement in the phase difference.

In order to extract this phase difference, the product of S ₁ and the conjugate complex S ₂ ^{~ is first} formed:
S ₁₂ = S ₁ .S ₂ ^~
= S ₀ ² .R ₁ (φ ₁ ) .R ₂ (φ ₂ ) ^~ .g (d ₁ ) .g (d ₂ ) .exp (j.4π. (D ₁ - d ₂ ) / λ) .A (φ ₁ ) .A (φ ₂ )

Only the phase factor exp (j.4π. (D ₁ - d ₂ ) / λ) is of interest for solving the problem, since this is where the vehicle movement is involved. The complex backscatter amplitudes R ₁ , R ₂ and the influence of the antenna pattern A (φ ₁ ), A (φ ₂ ) disturb, since the aspect angles φ ₁ and φ _{2 are} not known a priori.

Is the behavior of the backscattering phase independent of the aspect angle assumed for the stand goals, as z. B. is the case with triple mirrors, the following applies:
phase (R ₁ (φ ₁ ) .R ₂ (φ ₂ ) ^~ ) = ψ ₀ = const.

For the area of the main lobe, most antenna patterns have the property phase (A (φ ₁ )) = phase (A (φ ₂ )), so that it follows:
phase (A (φ ₁ ) .A (φ ₂ ) ^~ ) = 0.

This results in:
Δψ = phase (S ₁₂ ) = ψ ₀ + 4π. (D ₁ - d ₂ ) / λ

Since the value range is restricted to -π to + π in this arithmetic operation, a so-called "phase unwrapping" must be carried out, so that a continuous phase curve then results: Δψ _unwrapped .

The first target is at the coordinates (x _T1 , y _T1 ) and the second at (x _T2 , y _T2 ). The sensor position is described by (x, y).

It follows:
d ₁ = ((x _T1 - x) ² + (y - y _T1 ) ² ) ^1/2 ,
d ₂ = ((x _T2 - x) ² + (y - y _T2 ) ² ) ^1/2 .

If the targets are close to the road, it can be assumed that the measuring distances are not too small
(y - y _T1 ) ² <<(x - x _T1 ) ² and (y - Y _T2 ) ² <<(x - x _T2 ) ² ,
so that linear approximations can be used for the root expressions.

That makes:
d ₁ - d ₂ ≍ x _T1 - x _T2 + S. (y - y _T1 ) ² / (x _T1 - x) - S. (y - y _T2 ) ² / (x _T2 - x).

As mentioned initially, in the case of a non-linear movement, the Vehicle helix angle α (x) to errors in the correction angle estimate, the in turn is determined by α (x) = dy / dx.

The further procedure is therefore clear: the phase difference must be derived according to x and y '= dy / dx must finally be extracted from it.
d / dx (d ₁ - d ₂ ) = y '. ((y - y _T1 ) / (x _T1 - x) - (y - y _T2 ) / (x _T2 - x)) + S. ((y - y _T1 ) / (x _T1 - x)) ² - S. ((y - y _T2 ) / (x _T2 - x)) ²

Furthermore, it is assumed that only slight deviations from the ideal straight-ahead travel occur, which is synonymous with y ≍ 0.
d / dx (d ₁ - d ₂ ) ≍ -y '. (y _T1 / (x _T1 - x) - y _T2 / (x _T2 - x) + S. (y _T1 / (x _T1 - x)) ² - S. (y _T2 / (x _T2 - x)) ²

Overall, the following expression results for the vehicle helix angle α:
α = y '= - (d / dx (Δψ _unwrapped ) .λ / 4π - S. (y _T1 / (x _T1 - x)) ² + + S. (y _T2 / (x _T2 - x)) ² ) / ((y _T1 / (x _T1 - x) - y _T2 / (x _T2 - x))

For the further evaluation, the measured distances to the two target locations are used for the terms x _T1 - x and x _T2 - x: d _mess1 (n) and d _mess2 (n).

When deriving the phase difference, it should be noted that it does not exist as a function of the vehicle longitudinal coordinate x, but that the measurements (indexed with n) take place in cycles of the distance T: d / dx = 1 / (dx / dt) .d / dt = 1 / vd / dt = 1 / (v. T) .d / dn.
α (n) = - (d / dn (Δψ _unwrapped ) .λ / (4π.v (n) .T) - S. (y _T1 / d _mess1 ) ² + S. (y _T2 / d _mess2 ) ² ) / (y _T1 / d _mess1 - y _T2 / d _mess2 )

option A

The transverse positions y _T1 and y _T2 are known, e.g. B. on a reference line for quality monitoring of the factory sensor adjustment. The constellation y _T1 - y _T2 =: y _T should preferably be selected, as a result of which the phase fluctuations are minimized, so that the phase unwrapping can be carried out more easily.

It follows:
α (n) = - (d / dn (Δψ _unwrapped ) .λ / (4π.vT) - Sy _T ^2. ((1 / d _mess1 ) ² - (1 / d _mess2 ) ² )) / (y _T / d _mess1 + y _T / d _mess2 )

All quantities occurring in this equation are known or measured. The vehicle helix angle can thus be calculated for each measuring cycle, which can then be used to correct the measured transverse distances of the standing targets:
q _{mess_korr} (n) = q _mess (n) + α (n) .d _mess (n)

With q _{mess_korr} the regression is finally carried out. A correction angle is obtained for each target, the mean value of which can be used as the end result.

This method was applied to the measurement data in Fig. 4. In addition to this target, a second one was simulated, which was on the opposite side of the road (y _T1 = +1.5 m, y _T2 = -1.5 m) at a longitudinal distance of 10 m behind the first.

Fig. 5 shows the measured values 12, from which the course was 13 determines the vehicle skew angle with the phase difference method. Up to a distance of 60 m, the estimated and the specified vehicle helix angles agree very well, below this there are deviations, since the approximation of the large distances becomes worse here.

The transverse distances of the stationary targets were corrected with this angular profile 13 in order to compensate for the influence of the vehicle roll movement.

This results in the representation in FIG. 6. The method of linear regression can thus ultimately be applied to the measurement data 14 and a regression line 15 can be determined. In this case, the correction angle is 0.29 °.

Variant B

It is possible that there is no a priori knowledge about the location of the stand goals is. However, it is assumed that these are to the left and right of the lane lie in order to keep the phase fluctuations low.

Since a possible correction angle and the vehicle helix angle have the same effect on the transverse distances of all targets, the y coordinates of the targets are expressed in the formula for the vehicle helix angle α by the difference of their y positions:
y _T1 =: y ₀ + Δy and y _T2 =: y ₀ - Δy, y ₀ = center of gravity of the standing targets in the y direction.

The quadratic expressions are neglected. The linear terms are approximated as follows:
y _T1 / d _mess1 - y _T2 / d _mess2 ≍ (d _mess2 + d _mess1 ) / (d _mess1 .d _mess2 ) .Δy.

With Δy = mean over all measurements n of (q _mess1 (n) - q _mess2 (n)), only known quantities are then available:
α (n) = -d / dn (Δψ _unwrapped ) .λ / (4π.v (n) .T) .d _mess2 .d _mess1 / ((d _mess2 + d _mess1 ) .Δy).

For the simulation test of this variant, the previously described roll movement of the sensor vehicle was used. However, the stand target arrangement was varied:
y _T1 = +1.5 m, y _T2 = -3.0 m, Δx = 20 m.

When evaluating the simulated data, this information was given as assumed unknown.

Fig. 7 shows the measurement data 16 and the above-described manner from estimated vehicle slip angle 17, which is now already well differ at larger distances from the specified value. However, these deviations had only a minor influence on the end result: the linear regression resulted in a correction angle of 0.38 °. Reference number 1 vehicle
2 vehicle symmetry axis
3 item, fixed point for calibration, 3 a first item, 3 b second item
4 distance sensor axis
5 determined transverse distance values
6 best fit line
7 curved path, roll course, 7 a mean course of the path
8 first radiation lobe
9 second radiation beam
10 (noisy) measurement signal of the transverse distance
11 regression line due to noisy cross-distance signal
12 (noisy) measurement signal, vehicle helix angle
13 vehicle helix angle determined from the measurement signal
14 (noisy) measurement signal, trajectory
15 regression line
16 (noisy) measurement signal, vehicle helix angle
17 vehicle helix angle determined from the measurement signal

Claims (2)

1. Method for automatically correcting output values of a distance sensor in a vehicle ( 1 ) that is traveling on a path ( 7 ), which comprises the steps:
Emitting at least one electromagnetic pulse through the distance sensor,
Detecting an amplitude of an electromagnetic reflection signal reflected by an object ( 3 ), which was generated by the at least one electromagnetic pulse, as a function of an angle with respect to a distance sensor axis ( 4 ),
Determining a transverse distance (q) of the object from the path of the vehicle as a function of the transit time and the amplitude of the reflection signal and the angle,
Comparison of several transverse distance values ( 5 ) and
Determining a correction angle (δ _r ) for the distance sensor from the comparison result,
characterized in that
simultaneously a first reflection signal (3 a) and a second reflection signal is detected (3 b) of a second article from a first article and a first transverse distance value and a second transverse distance value is determined, wherein the first object and the second object to the vehicle at an angle include, a phase difference between the first and the second reflection signal is determined and the correction angle (δ _r ) for the distance sensor is determined from the phase difference. 2. The method according to claim 1, characterized in that the path ( 7 ) of the vehicle is subsequently calculated from the time course of the phase difference.