Method for checking a vehicle dynamics model

Abstract

A method for checking a vehicle dynamics model of a vehicle, with which a value of an output variable is ascertainable from a value of a variable input variable, for multiple values of the input variable respectively associated model-based values of the output variable being ascertained with the aid of vehicle dynamics model, for the multiple values of the input variable at the vehicle respectively associated vehicle-based values of the output variable being ascertained, difference values being ascertained from mutually corresponding model-based values and vehicle-based values, respectively, an updated dataset of the ascertained difference values being compared with a comparison dataset with the aid of a comparison method and a concordance measure being ascertained in the process, and the vehicle dynamics model being determined to be valid if the concordance measure meets a predefined concordance criterion.

Claims

1. A method for checking a vehicle dynamics model of a vehicle, with which a value of an output variable is ascertainable from a value of a variable input variable, the method being executed on a processing unit installed in the vehicle which is coupled to a steering angle sensor of the vehicle and to a yaw-rate sensor of the vehicle, and the method comprising the following steps: during operation of the vehicle as the vehicle is driving and in real-time, performing the following steps: detecting using the steering angle sensor of the vehicle, a plurality of steering angles of the vehicle; for multiple values of the input variable, ascertaining respectively associated model-based values of the output variable using the vehicle dynamics model, wherein the multiple values of the input variable are the plurality of steering angles detected by the steering angle sensor; for the multiple values of the input variable at the vehicle, ascertaining, in parallel with the detecting using the steering angle sensor, respectively associated vehicle-based values of the output variable, wherein the vehicle-based values of the output variable are ascertained using the yaw-rate sensor of the vehicle; respectively ascertaining difference values for mutually corresponding ones of the model-based values and the vehicle-based values; comparing an updated dataset of the ascertained difference values with a comparison dataset using a comparison method and ascertaining a concordance measure based on the comparing; determining the vehicle dynamics model to be valid based on the concordance measure meeting a predefined concordance criterion; and based on determining the vehicle dynamics model to be valid, using the vehicle dynamics model by a driver assistance function of the vehicle; wherein the vehicle dynamics model predicts a behavior of the vehicle based on the plurality of detected steering angles.

2. The method as recited in claim 1, wherein the difference values are ascertained repeatedly in temporal succession at predetermined time intervals, during operation of the vehicle while the vehicle is driving.

3. The method as recited in claim 2, wherein the updated dataset is determined only from a predetermined number of the difference values.

4. The method as recited in claim 1, wherein a histogram difference is used as the comparison method.

5. The method as recited in claim 1, wherein a Jensen-Shannon divergence is used as a comparison method.

6. The method as recited in claim 1, wherein the vehicle dynamics model is determined to be invalid when the concordance measure does not meet the predefined concordance criterion, and an error response is then initiated.

7. The method as recited in claim 1, wherein the vehicle dynamics model is used for vehicle functions based on the vehicle dynamics model being determined to be valid.

8. A processing unit configured to check a vehicle dynamics model of a vehicle, with which a value of an output variable is ascertainable from a value of a variable input variable, the processing unit, when installed in the vehicle and coupled to a steering angle sensor of the vehicle and a yaw-rate sensor of the vehicle, configured to: during operation of the vehicle as the vehicle is driving and in real-time, the processor being configured to: detect using the steering angle sensor of the vehicle, a plurality of steering angles of the vehicle; for multiple values of the input variable, ascertain respectively associated model-based values of the output variable using the vehicle dynamics model, wherein the multiple values of the input variable are the plurality of steering angles detected by the steering angle sensor; for the multiple values of the input variable at the vehicle, ascertain, parallel with the detection using the steering angle sensor, respectively associated vehicle-based values of the output variable, wherein the vehicle-based values of the output variable are ascertained using the yaw-rate sensor of the vehicle; respective ascertain difference values for mutually corresponding ones of the model-based values and the vehicle-based values; compare an updated dataset of the ascertained difference values with a comparison dataset using a comparison method and ascertaining a concordance measure based on the comparing; determine the vehicle dynamics model to be valid based on the concordance measure meeting a predefined concordance criterion; based on determining the vehicle dynamics model to be valid, using the vehicle dynamics model by a driver assistance function of the vehicle; and wherein the vehicle dynamics model predicts a behavior of the vehicle based on the plurality of detected steering angles.

9. A non-transitory machine-readable memory medium on which is stored a computer program for checking a vehicle dynamics model of a vehicle, with which a value of an output variable is ascertainable from a value of a variable input variable, the computer program, when executed by a computer installed in the vehicle and coupled to a steering angle sensor of the vehicle and a yaw-rate angle-sensor of the vehicle, causing the computer to perform the following steps: during operation of the vehicle as the vehicle is driving and in real-time, performing the following steps: detecting using the steering angle sensor of the vehicle, a plurality of steering angles of the vehicle; for multiple values of the input variable, ascertaining respectively associated model-based values of the output variable using the vehicle dynamics model, wherein the multiple values of the input variable are the plurality of steering angles detected by the steering angle sensor; for the multiple values of the input variable at the vehicle, ascertaining, in parallel with the detecting using the steering angle sensor, respectively associated vehicle-based values of the output variable, wherein the vehicle-based values of the output variable are ascertained using the yaw-rate sensor of the vehicle; respectively ascertaining difference values for mutually corresponding ones of the model-based values and the vehicle-based values; comparing an updated dataset of the ascertained difference values with a comparison dataset using a comparison method and ascertaining a concordance measure based on the comparing; determining the vehicle dynamics model to be valid based on the concordance measure meeting a predefined concordance criterion; based on determining the vehicle dynamics model to be valid, using the vehicle dynamics model by a driver assistance function of the vehicle; and wherein the vehicle dynamics model predicts a behavior of the vehicle based on the plurality of detected steering angles.

Description

BRIEF DESCRIPTION OF THE DRAWINGS

(1) FIG. 1 schematically shows a vehicle, in which an example method according to the present invention is implementable.

(2) FIG. 2 schematically shows a sequence of an example method according to the present invention in one preferred specific embodiment.

(3) FIGS. 3 through 5 show diagrams for explaining the example method shown in FIG. 2.

DETAILED DESCRIPTION OF EXAMPLE EMBODIMENTS

(4) FIG. 1 schematically depicts a vehicle 100, in which a method according to the present invention is implementable. A processing unit 150 on which, for example, a driver assistance system may be operated, which takes a yaw rate into consideration, is provided, for example, in vehicle 100.

(5) For this purpose, a yaw rate sensor 155 is also provided, with the aid of which a yaw rate may be ascertained and conveyed to processing unit 150. A steering wheel 140 is also indicated, with the aid of which a steering angle may be set. The steering angle may be detected, for example, with the aid of steering angle sensor 145 and conveyed to processing unit 150. It is understood that a steering angle may be set or changed also in an automated manner in conjunction with a driver assistance function and/or in conjunction with autonomous or automated driving.

(6) FIG. 2 schematically depicts a sequence of a method according to the present invention in one specific embodiment. This method may, for example be carried out on processing unit 150, as it is shown in FIG. 1. FIGS. 3 through 5 show diagrams for explaining the method shown in FIG. 2.

(7) For this purpose, values W.sub.E are each detected or ascertained repeatedly or in temporal succession for one or for multiple input variables E. The input variable may, for example, be a steering angle which, or the values of which, may be ascertained as explained with reference to FIG. 1.

(8) Based on an instantaneous value W.sub.E of the steering angle, a corresponding, vehicle-based value W.sub.F of an output variable A is then ascertained at the vehicle. The output variable A may, for example, be a yaw rate which, or the values of which, may be ascertained or measured with the aid of the yaw rate sensor as explained with reference to FIG. 1.

(9) Based on instantaneous value W.sub.E of the steering angle, a corresponding, model-based value W.sub.M of output variable A, i.e., in this case a yaw rate, is ascertained or calculated in parallel thereto with the aid of a vehicle dynamics model M.

(10) Such a vehicle dynamics model, with which a yaw rate {dot over (ψ)} may be ascertained based on a steering angle δ, may be represented by the following differential equations:

(11) $\left[\begin{array}{c}\stackrel{\mathrm{.Math.}}{\psi }\\ \stackrel{.}{\beta }\end{array}\right]=\left[\begin{array}{cc}-\frac{1}{v}\frac{c_f{l}_v^2+c_r{l}_r^2}{\Theta }& -\frac{c_f{l}_f+c_r{l}_r}{\Theta }\\ -1-\frac{1}{v^2}\frac{c_f{l}_f+c_r{l}_r}{m}& -\frac{1}{v}\frac{c_f+c_r}{m}\end{array}\right]\left[\begin{array}{c}\stackrel{.}{\psi }\\ \beta \end{array}\right]+\left[\begin{array}{c}\frac{c_f+c_f}{\Theta }\\ \frac{1}{v}\frac{c_f}{m}\end{array}\right]\frac{\delta }{i}$

(12) In these equations, ß indicates a lateral slip angle, δ indicates the steering angle, i indicates the ratio between the steering angle (or steering wheel angle) and the wheel angle, v indicates the vehicle velocity, θ indicates the yaw inertia, c.sub.f and c.sub.r indicate the front and rear lateral stability of the tires, l.sub.f and l.sub.r indicate the front and rear distance between the wheel and center of gravity, and m indicates the vehicle mass. The wheel angle in this case is thus determined as a quotient of the ratio i and the steering angle δ.

(13) The values for these variables—with the exception of input variable δ and vehicle velocity v—are in this case vehicle-specific parameters and are generally known or may be measured or calculated.

(14) For this purpose, an exemplary profile of a steering angle is depicted in the upper diagram in FIG. 3 as input variable E over time t in seconds. A corresponding profile V.sub.1 for the actual or measured profile of the yaw rate as output variable A, as well as a profile V.sub.2 for the profile of the yaw rate ascertained with the aid of the vehicle dynamics model, are depicted in the lower diagram. The values W.sub.F and W.sub.M may be ascertained repeatedly, for example, at an interval of one second or 100 ms, respectively.

(15) A difference value W.sub.D is also ascertained or formed from respectively one pair of mutually corresponding values W.sub.F and W.sub.M. This difference value W.sub.D is then fed to a buffer memory B, in which all difference values ascertained in this way (in temporal succession) are stored.

(16) An updated dataset H.sub.A is then formed from these difference values W.sub.D present in buffer memory B. For this purpose, a particular number, for example, 300, the most updated or the latest difference values ascertained and stored in the buffer memory may be used, for example. In this context, it is also conceivable that only this number of values is stored or may be stored in the buffer memory, older values on the other hand are deleted or overwritten.

(17) This updated dataset H.sub.A is then processed in the form of a histogram, i.e., of a frequency distribution. This updated dataset may then be compared in conjunction with a comparison method or a comparison step 210 with a comparison dataset or reference dataset H.sub.R— also depicted in the form of a histogram. In this case, a concordance measure ΔH is ascertained and then compared with a predefined threshold value ΔH.sub.s.

(18) For this purpose, FIG. 4 depicts a comparison dataset H.sub.R on the left and an updated dataset H.sub.A on the right, each in the form of a histogram. Here, a number N of difference values is plotted over a value of difference values W.sub.D in 10.sup.−2 in each case. For comparing, the two histograms must, if necessary, still be standardized.

(19) The comparison method is explained briefly and by way of example below with reference to the previously mentioned Jensen-Shannon divergence. This divergence is based on the so-called Kullback-Leibler divergence. The Kullback-Leibler divergence measures to what degree a probability distribution P(x) differs from a second probability distribution Q(x). The Kullback-Leibler divergence D.sub.KL for discrete probability distributions in the same probability space is defined in this case as follows:

(20) $D_{\mathrm{KL}}\left(P||Q\right)=\underset{x\in X}{\overset{}{.Math.}}P\left(x\right)\log \left(\frac{P\left(x\right)}{Q\left(x\right)}\right)$

(21) The Kullback-Leibler divergence is always not negative, i.e., D.sub.KL(P∥Q)≥0, but neither finite nor symmetrical. In practice, therefore, the sum of the two Kullback-Leibler divergences
D.sub.KL2(P∥Q)=D.sub.KL2(Q∥P)=D.sub.KL(P∥Q)+D.sub.KL(P∥Q)
is often applied in order to ensure the symmetry. However, the infinite values of the Kullback-Leibler divergence result in problems in the practical implementation. For this reason, the so-called Jensen-Shannon divergence D.sub.JS is preferred, which is symmetrical and limited and is based on the Kullback-Leibler divergence as follows:
D.sub.JS(P∥Q)=½D.sub.KL(P∥M)+½D.sub.KL(Q∥M)
where M=0.5.Math.(P+Q) applies. The result of a comparison of updated dataset H.sub.A and comparison dataset H.sub.R, which represent the histograms or frequency distributions P and Q in the above notation, is a concordance measure (or divergence measure) ΔH, in the case of the Jensen-Shannon divergence, a positive number.

(22) FIG. 5 shows a profile of such a concordance measure (or divergence measure) ΔH over time t in seconds, as results for the profiles of the output variables in the lower diagram of FIG. 3. In this case, it is clearly apparent that the concordance measure ΔH increases sharply, in particular in the area of the unusual and sharp steering angle changes between the points in time 70 and 80 seconds.

(23) In the present case, a value of 0.25, for example, may be selected as threshold value ΔH.sub.s, the vehicle dynamics model being determined to be valid if the concordance measure is below the threshold value, i.e., if: ΔH<ΔH.sub.s applies.

(24) In a step 220, an error response may then be initiated if the vehicle dynamics model is not determined to be valid or is determined to be invalid, as previously explained in detail above. If, however, the vehicle dynamics model is determined to be valid, it may be used as usual for desired functions.

(25) As is apparent from FIG. 5, it may therefore also happen that the vehicle dynamics model is determined to be only temporarily invalid and is then not used during this time period. Later, however, it may then be used again, for example.