WO2020148099A1

WO2020148099A1 - Method for providing an integrity range of a parameter estimation

Info

Publication number: WO2020148099A1
Application number: PCT/EP2020/050019
Authority: WO
Inventors: Thomas FRIEDERICHS; Jens Strobel; Michael Baus; Marco Limberger; Thomas Ulrich; Sebastian ROITH; Mohamed Ben Tahar
Original assignee: Robert Bosch Gmbh
Priority date: 2019-01-16
Filing date: 2020-01-02
Publication date: 2020-07-23
Also published as: KR20210108484A; DE102019200423A1; CN113302452A

Abstract

The invention relates to a method for providing an integrity range (1, 1) of a parameter estimation, the integrity range describing the range in which an estimated parameter lies within a minimum probability. The method comprises at least the following steps: a) acquiring at least one piece of integrity information, b) evaluating the at least one piece of integrity information acquired in step a), and c) providing the integrity range by outputting at least two pieces of information (2, 3, 12, 13) that describe a non-rotation-invariant shape (5, 15).

Description

description

title

Method for providing an integrity area of a parameter estimate

The invention relates to a method for providing an integrity area of a parameter estimate, a computer program, a machine-readable storage medium and a control device for a motor vehicle. The invention is particularly suitable for use in autonomous driving.

State of the art

One of the most important challenges for autonomous driving is the most accurate and reliable determination of the autonomous vehicle's own position. The autonomous vehicle usually has sensors, such as inertial sensors, wheel sensors,

Environment sensors, GNSS sensors, optical and / or acoustic sensors, by means of which the vehicle can estimate its own position. In this context, it is helpful if information about its (expected) estimation accuracy is also output for a determined own position. In this context, for example, the confidence of the determined own position can be represented by a so-called "Protection Level" (short: "PL"). The PL can describe a statistical error limit, the calculation of which is generally based on statistical considerations and, if necessary, additionally on a suitable coordination of the estimation algorithms.

The concept of providing the protection level is particularly widespread in aviation. However, the solutions developed are based on the

Application area of autonomous driving is not easily transferable. In particular, house canyons and their influence on satellite signals represent problems that do not occur in aviation applications. In addition, the orientation of the Aircraft (in particular the orientation of the aircraft in a global coordinate system) generally does not matter for the location of the aircraft. Improved methods for calculating a protection level that is as reliable as possible are therefore desirable, which can deliver reliable results especially in difficult environments, such as in urban areas.

Disclosure of the invention

Here, according to claim 1, a method for providing an integrity area of a parameter estimate is proposed, the integrity area describing the area in which an estimated parameter with a

The minimum probability is, the method comprising at least the following steps:

a) determining at least one piece of integrity information,

b) evaluating the at least one determined in step a)

Integrity information,

c) Providing the integrity area by outputting at least two pieces of information that describe a non-rotation-invariant form.

Steps a), b) and c) are carried out in particular in the order given. The solution proposed here advantageously allows the integrity area to be made available depending on the direction, in particular depending on or in relation to the orientation or orientation of a motor vehicle (for which, for or in which the parameter estimation is carried out). The alignment or orientation here means in particular the alignment and / or orientation in a global coordinate system. This is particularly noticeable

Improvement over an only circular one

Integrity area.

The integrity area describes the area in which an estimated

Parameter (value) with a minimum probability (actually). The estimated parameter (value) basically describes an (individual, in particular momentary) estimation result of the parameter estimation. In other words, this means in particular that the integrity area describes the area in which a real or actual value of an estimated one Parameter with a minimum probability. Such a

Integrity area can also be referred to as the so-called “protection level”.

The minimum probability is usually a predefined minimum probability. The is preferably

Minimum probability 90%, particularly preferably 95% or even 99%.

The range of integrity is preferably a protection level. The protection level usually describes the (spatial, in particular two- or three-dimensional) area in which an estimated parameter (value) with a minimum probability (actually) lies. The estimated parameter (value) basically describes an (individual, in particular momentary) estimation result of the parameter estimation. In other words, this means in particular that the protection level describes the area in which a real or actual value of an estimated parameter with a

The minimum probability is.

In other words, a protection level describes in particular a confidence interval or a (spatial) confidence range, in which the true value of an estimated parameter can be compared with a

Minimum probability. The estimated value of the parameter is usually in the middle or the center of the confidence interval or confidence range.

The minimum probability with which a real or actual value of an estimated parameter is actually in a protection level is still much higher than with "usual" integrity areas. The minimum probability here is usually over 99.99%, particularly preferably over 99.999% or even over 99.9999%. The minimum probability at the protection level can also not be expressed in percent but in possible errors in a certain time interval. A protection level can, for example, be defined so that the parameter in question is outside the protection level at most once every 10 years. The protection level can be, for example, either as a unitless probability or as a rate, i.e. as

Error occurrence probability over a time interval can be expressed. In step a), at least one piece of integrity information is ascertained, in particular via the parameter estimate or an estimated one

Parameter. For example, a covariance matrix or at least one value from a covariance matrix for the parameter estimation and / or from the parameter estimation can be determined. In this context, it is particularly conceivable that values are read from a covariance matrix. The covariance matrix can be determined, for example, by a filter model, such as a Kalman filter, and in this connection can result, for example, from the parameter estimation carried out using the filter model. Sensor data of the sensors of the motor vehicle, which are also mentioned here, can be supplied to the filter model as input data. As an alternative or supplement to the filter model, state-describing functions are fundamentally conceivable, in particular so-called “state observers”, for example extensive fuzzy methods or the like.

In step b), the at least one piece of integrity information determined in step a) is evaluated. For this purpose, for example, mathematical methods for evaluating a matrix, such as a covariance matrix, can be used.

In step c) the integrity area is provided by outputting at least two pieces of information that describe a non-rotationally invariant (geometric) shape. The non-rotationally invariant (geometric) shape can be, for example, an ellipse, a rectangle, a diamond or the like. It is preferably an ellipse. If necessary, a rectangle can be derived from the ellipse. The at least two pieces of information may e.g. are (support) vectors that span and / or (unambiguously) define the non-rotationally invariant (geometric) shape. If the integrity area is to be determined three-dimensionally, the ellipse can become an ellipsoid and the rectangle a cuboid. In the case of a multidimensional representation, a hyperellipsoid is also conceivable, for example.

According to an advantageous embodiment, it is proposed that the estimated parameter is a driving operation parameter of a motor vehicle. The driving operation parameter is usually a safety-critical or safety-relevant parameter of the ferry operation a motor vehicle. It is preferably the

Driving operation parameters by a (safety-critical or

safety-relevant) parameters of the ferry operation of an at least partially automated or even autonomously operating (or operated)

Motor vehicle.

A driving operation parameter is used here in particular

Understand parameters that contribute to the spatial ferry operation of a motor vehicle or the operation of a motor vehicle in space

describe. In particular, the driving operation parameter helps at least to describe an own movement and / or own position of a motor vehicle. The driving operation parameter can be, for example, an (own) position, an (own) speed, (own) acceleration or a position (or orientation) of the motor vehicle. The driving operation parameter is preferably a self-position of the motor vehicle.

The range of integrity is also preferably determined in real time or

provided. In this context, the area of integrity

for example, determined and / or provided while the motor vehicle is traveling.

The method is preferably used to provide an integrity area that describes the integrity of an estimate of a vehicle's own position. In other words, this means in particular that the parameter is preferably an own position of a vehicle. The method can (thus) for example provide a

Integrity range serve a position estimate of a vehicle position. The integrity area can describe the area in which an estimated own position of a vehicle with a minimum probability (actually) lies. Alternatively or cumulatively to the estimation of the vehicle's own position, the method can also be used to estimate the

Own speed, orientation, own movement or the like of the vehicle can be used.

According to a further advantageous embodiment, it is proposed that in step a) the at least one piece of integrity information based on data from at least one sensor is determined, which is preferably arranged in or on a motor vehicle. The determination of the

Integrity information based at least also on GNSS (global navigation satellite system) data (and for example additional GNSS correction data or data comprising both GNSS position data and GNSS correction data) of a GNSS sensor of a motor vehicle. Alternatively or cumulatively, the integrity information can be determined at least on the basis of data from an environment sensor of a motor vehicle. The environment sensor can be, for example, a camera, a RADAR sensor, a LIDAR sensor and / or an ultrasonic sensor.

According to a further advantageous embodiment, it is proposed that a covariance matrix be evaluated in step b). The covariance matrix usually contains the uncertainties of the state parameters determined by the system, as well as their correlations with each other. These uncertainties (variances and covariances) form the so-called "second central moment in statistics" and represent the individual matrix elements. The resulting one

Covariance matrix is symmetrical and has as many rows as there are columns. In the case of a two-dimensional position determination, the

Covariance matrix also two-dimensional. That means it has two rows and two columns (a total of 4 matrix elements). The input parameters for the matrix can be, for example, the uncertainties of the state parameters themselves and / or the correlations with one another, that is to say uncertainties between the state parameters (= relationship measure for the relationship between two state parameters).

The covariance is preferably evaluated such that an ellipse is determined from it. Using mathematical methods, such as solving the eigenvalue problem of the matrix and setting up the characteristic polynomial of the matrix and zeroing this polynomial, the

Determine the eigenvalues of the matrix. The individual matrix elements are used as input parameters for this. The eigenvalues then calculated represent the length of the two semiaxes of an ellipse, which initially only represents a pure error ellipse. The associated eigenvectors of the matrix can be determined for the specific eigenvalues of the matrix. Using this

Eigenvectors can determine an angle that is the orientation of the ellipse within a defined coordinate system. So you can first evaluate the following output parameters from the covariance matrix: The eigenvalues of the matrix, the eigenvectors associated with the eigenvalues and / or the orientation angle with respect to a fixedly defined one

Koo rd i n aten syste m.

Apart from the center point of the ellipse (which can preferably be placed here in the center point or center of gravity of the motor vehicle), you now have all the parameters for uniquely describing an ellipse. Up to this point it is a so-called error ellipse. This error ellipse can now be inflated / enlarged using a factor (scalar) until a desired confidence interval or confidence interval is reached. This means that this factor is applied to the lengths of the ellipse half axes. The one from it

resulting ellipse maintains its orientation. Only the semiaxes are extended accordingly. One now speaks of a confidence ellipse. The factor by which it is “inflated” can be determined, for example, by means of quantile tables from the associated statistical probability distributions.

According to a further advantageous embodiment, it is proposed that the at least two pieces of information output in step c) form an ellipse

describe. As stated above, this at least two pieces of information can be, for example, two eigenvectors. The ellipse can be arranged so that the estimated parameter lies in its center. Furthermore, the ellipse can be oriented so that it lies in a horizontal plane.

The integrity area is preferably a horizontal integrity area. In other words, this means in particular that the area of integrity relates to an area lying in a horizontal plane. Accordingly, the ellipse can also be an ellipse that lies in a horizontal plane.

On the basis of the ellipse possibly determined as an intermediate step, a further or different non-rotation-invariant shape can also be determined, which is then provided in step c). For example, a box or a rectangle can be determined that surrounds or envelops the ellipse. Here it is in particular a rectangle with the smallest possible area, but which completely surrounds the ellipse.

According to a further advantageous embodiment, it is proposed that in step c) at least one further item of information is output, which is an orientation angle of the non-rotation-invariant shape with respect to a

Motor vehicle coordinate system describes. In this context, the orientation angle can be based in particular on the eigenvectors of the

Covariance matrix can be determined. In this context in particular, it is particularly advantageous if the non-rotationally invariant form is in one

Motor vehicle coordinate system is transformed.

According to a further advantageous embodiment, it is proposed that a confidence area be provided as the integrity area. The confidence range can be provided, for example, by scaling the integrity range. The (geometric) shape provided in step c) generally describes and / or surrounds the area of integrity. This (geometric) shape can also be scaled (in particular enlarged), for example by

Increase the minimum probability or set it to a certain value. The range of integrity thus scaled then forms one

Confidence area. In particular, as described above, a confidence ellipse can be provided in this way.

According to a further aspect, a computer program for performing a method presented here is proposed. In other words, this relates in particular to a computer program (product) comprising commands which, when the program is executed by a computer, cause the computer to carry out a method described here.

According to a further aspect, a machine-readable storage medium is proposed on which the computer program proposed here

is saved. The machine-readable storage medium is usually a computer-readable data carrier.

According to a further aspect, a control device for a motor vehicle is proposed, the control device for carrying out a here

proposed procedure is established. For this purpose, the control unit can Have memory on which a program for executing the method is stored. Furthermore, the control device can have a processor that can access the memory and execute the program.

The details, features and advantageous refinements discussed in connection with the method can accordingly also apply to the computer program presented here, the storage medium and / or the

Control unit occur and vice versa. In this respect, the ones there

Comments on the detailed characterization of the features are fully referenced.

The solution presented here and its technical environment

explained in more detail below with reference to the figures. It should be pointed out that the invention is not intended to be limited by the exemplary embodiments shown. In particular, unless explicitly stated otherwise, it is also possible to extract partial aspects of the facts explained in the figures and to combine them with other components and / or knowledge from other figures and / or the present description. It shows schematically:

1: a sequence of a method proposed here,

2: an exemplary illustration of what is proposed here

Process,

3: a motor vehicle with a control unit proposed here.

1 schematically shows a sequence of a method proposed here. The method serves to provide an integrity area 1, 11 of a parameter estimate, the integrity area describing the area in which an estimated parameter with a minimum probability lies.

The sequence of steps a), b) and c) shown with blocks 110, 120 and 130 generally arises in a normal operating sequence. In block 110, at least one piece of integrity information is determined in accordance with step a). In block 120, the at least one piece of integrity information determined in step a) is evaluated in accordance with step b). In block 130, according to step c), the integrity area is provided by outputting at least two Information 2, 3, 12, 13, which describe a non-rotationally invariant form 5, 15.

2 schematically shows an exemplary illustration of the method proposed here. The reference symbols are used uniformly, so that reference can be made in full to the preceding explanations, in particular for FIG. 1.

An integrity area 1 around a motor vehicle 30 is shown in FIG. 2. This integrity area 1 has, for example, the shape of an ellipse 5. This ellipse 5 lies here, for example, in a horizontal plane. At least two pieces of information 2, 3 are available for clamping the ellipse 5. The information 2 is here, for example, the main axis of the ellipse 5 and the information 3 is, for example, the minor axis 3 of the ellipse 5.

Furthermore, an orientation angle 4 of the ellipse 5, in particular the main axis 2 of the ellipse 5, relative to a motor vehicle coordinate system 31 of the motor vehicle 30 is determined and provided as information. The non-rotationally invariant shape of the integrity area 1 thus allows it to be made available as a function of or in relation to the direction of travel 32 of the motor vehicle 30. This can contribute to improving the determination of the own position of motor vehicle 30, in particular when cornering. This represents, in particular, a noticeable improvement over areas of integrity which only describe a circle 40 over a radius 41.

FIG. 2 also illustrates a possible, further non-rotationally invariant form that describes the integrity area 11. This is a box 15. The box 15 can be described by way of example using two pieces of information 12, 13. The information 12 exemplifies the (maximum) deviation along the direction of travel 32 and the information 13 exemplifies the (maximum) deviation transverse to the direction of travel 32.

3 schematically shows a motor vehicle 30 with a control device 20 proposed here. The control device 20 is set up to carry out a method presented here. The motor vehicle 30 also has a sensor 33 which can transmit data to the control device 20. In other words, the solution presented here can be summarized in particular as follows: The so-called protection level is a critical factor for safety concepts in the field of autonomous driving.

The protection level describes a confidence range in which the actual position of an object whose position is to be determined is very likely. The possibility that the actual position is outside the protection level is extremely low. This high level of safety is essential for applications in the field of autonomous driving.

This determined presumed position is usually in the center of the protection level. For applications in the area (e.g. the

Position determination in one level) the protection level is usually described as a circle in one level. A circle as a protection level has the advantage that it can be described with just one parameter (the radius). However, such a protection level is not very precise. The circle is just a very rough assumption of the uncertainty of the particular position. In fact, based on the available data, much more precise assessments are often possible in certain directions in the plane. The directions in which these more precise assessments are possible often coincide with the directions in which exact vehicle positions must be known for applications in autonomous driving. For example, it is very important for autonomous driving to know exactly what uncertainty exists when determining the position perpendicular to the direction of travel. An uncertainty in determining the position parallel to the direction of travel, on the other hand, is often more acceptable. In fact, the available data often make it more precise in the direction perpendicular to the direction of travel

To ensure position determination. This is ensured, for example, on the basis of road markings which are particularly precise

Enable position detection if the position of the lane markings itself is known. In such cases, the actual accuracy of the position determination cannot be represented with a classic, circular protection level for technical reasons. The solution has now been described to map these uncertainties much better with an elliptical protection level. This is particularly relevant for safety aspects in the field of highly automated or autonomous driving. In addition to the shape of the protection level (ellipse), properties such as the spatial direction can also be taken into account in a coordinate system. In this context, the protection level interval should be in another

Coordinate system, for example in a body coordinate system of the

Motor vehicle be specified. A transformation of the protection level between a vehicle coordinate system and a global one

Coordinate system is possible with a matrix transformation. This approach is not possible with the classic definition of the protection level (specified with a radius), for example used in the aerospace sector, and is also not necessary because it is circular or spherical

Integrity area can not be aligned.

As described, the actual uncertainty of positioning is usually better with the shape of an in alignment

Vehicle coordinate system arranged ellipse described. Then uncertainties in different spatial directions can be mapped well relative to the vehicle. The protection level described here can therefore differentiate between a motor vehicle longitudinal direction uncertainty and a transverse direction uncertainty transverse to the vehicle longitudinal direction. In particular, when a motor vehicle is cornering, a further rotation can take place relative to the vehicle coordinate system, for example the

Curve radius considered.

All - apart from one or more scaling factors - necessary parameters to describe the ellipse of the protection level can be from one

Covariance matrix with components in a diagonal axis of the matrix as well as further components linking these components can be found in the further cutouts of the matrix. In addition to one or more scaling factors, the components in the diagonal of the matrix primarily describe the extent of the elliptical protection level. The other components in the cutouts of the matrix describe in particular the rotation of the ellipse. The matrix can transform into a

Motor vehicle coordinate system are converted so that only the components in the axis diagonal exist (then called main components), the other components of the matrix becoming zero.

Claims

Expectations

1. Method for providing an integrity area (1, 11) of a

Parameter estimation, wherein the integrity range describes the range in which an estimated parameter with a minimum probability lies, the method comprising at least the following steps:

a) determining at least one piece of integrity information,

b) evaluating the at least one determined in step a)

Integrity information,

c) Providing the integrity area by outputting at least two pieces of information (2, 3, 12, 13) that describe a non-rotationally invariant form (5, 15).

2. The method of claim 1, wherein the estimated parameter is a driving operation parameter of a motor vehicle (30).

3. The method according to claim 1 or 2, wherein in step a) the at least one integrity information is determined on the basis of data from at least one sensor (33), which is preferably arranged in or on a motor vehicle (30).

4. The method according to any one of the preceding claims, wherein a covariance matrix is evaluated in step b).

5. The method according to any one of the preceding claims, wherein the at least two items of information output in step c) describe an ellipse (5).

6. The method according to any one of the preceding claims, wherein in step c) at least one additional information is output, the one

Describes the orientation angle (4) of the non-rotationally invariant form (5) in relation to a motor vehicle coordinate system (31).

7. The method according to any one of the preceding claims, wherein as

Integrity area (1, 11) a confidence area is provided.

8. Computer program for performing a method according to one of claims 1 to 7.

9. Machine-readable storage medium on which the computer program according to claim 8 is stored.

10. Control device (20) for a motor vehicle (30), wherein the control device (20) is set up to carry out a method according to one of claims 1 to 7.