Method for checking a vehicle dynamic model

Abstract

Method for checking a vehicle dynamics model of a vehicle, with which method a value of an output variable can be determined from values of variable input variables, wherein for a plurality of values of the input variables a respectively assigned model-based value of the output variable is determined by means of the vehicle dynamics model, wherein for a plurality of values of the input variables a respectively assigned vehicle-based value of the output variable is determined on the vehicle, wherein a respectively assigned difference value is determined from the model-based value and the vehicle-based value which correspond to one another, wherein a current data set of the determined difference values is compared with a comparison data set by means of a comparison method, and in which a conformity measure is determined, and wherein the vehicle dynamics model is determined to be valid if the conformity measure satisfies a predefined conformity criterion.

Description

Technical Field

[0001] This invention relates to a method for inspecting a vehicle dynamic model, which allows the determination of output parameter values ​​from variable input parameter values, and to a computing unit and a computer program for implementing the method. Background Technology

[0002] Vehicle dynamics models are playing an increasingly important role, particularly in the field of driver assistance systems, and also generally in the field of autonomous or automated driving of vehicles, especially buses or trucks.

[0003] A vehicle dynamics model is a model of a vehicle that can represent or depict its dynamic operation, allowing the determination of output parameter values ​​from variable input parameters. One example is determining the composite yaw rate for a specific steering angle. This enables the pre-calculation of vehicle performance, at least over a certain time period. Summary of the Invention

[0004] According to the present invention, a method for inspecting a vehicle dynamic model, a computing unit for implementing the method, and a computer program are provided. Advantageous designs are the subject of the dependent claims and the following description.

[0005] This invention relates to a method for inspecting a vehicle dynamics model, which allows the determination of the value of at least one output parameter from the values ​​of at least one variable input parameter. It goes without saying that, depending on the type of vehicle dynamics model, the corresponding values ​​of one or more output parameters can be determined from the values ​​of one or more input parameters.

[0006] Using such vehicle dynamics models, it has been possible, to date, to acquire and, if necessary, validate different vehicle design schemes before the vehicle's first operation, especially within a so-called homogeneous range. The results obtained using vehicle dynamics models are then evaluated visually, and therefore particularly by professionals.

[0007] Further optimization allows such models to be used for pre-calculation. If multiple trajectories (possible paths) are calculated comprehensively and compared within the avoidance strategy, then such dynamic models can be used, for example, during autonomous driving. Here, it is particularly important to pre-calculate how the vehicle will behave when different options are chosen, only within a short timeframe. Another case involves, for example, energy-optimal cornering.

[0008] Such vehicle dynamics models are equally important for driver assistance systems, but especially in the realm of autonomous or automated driving. As already shown, there is another crucial point: these vehicle dynamics models must also consistently and accurately depict or represent the vehicle's actual performance. Otherwise, the various functions of the vehicle—particularly in terms of safety—may no longer be possible or guaranteed.

[0009] It's important to consider that the actual condition or performance of a vehicle may change over time, for example, due to wear and tear on different components. However, it's also possible to consider that the vehicle dynamics model may not accurately depict the vehicle under certain, temporary conditions. Besides wear and tear, modifications to the vehicle, such as installing a roof box or bicycle rack, may also have an effect.

[0010] In the proposed method, the vehicle dynamics model is now examined and evaluated to determine whether it can be considered valid, that is, whether the vehicle dynamics model accurately depicts the actual performance of the vehicle and is therefore (currently or continuing to be) usable or permitted to be used.

[0011] Therefore, for multiple values ​​of input parameters, such as steering angle, the respective model-based values ​​of the output parameters, such as yaw rate, are obtained using the vehicle dynamics model. Furthermore—particularly in parallel—the respective vehicle-based values ​​of the output parameters are obtained on the vehicle for multiple values ​​(i.e., identical values) of the input parameters. This involves values ​​obtained by other means, i.e., not by means of the vehicle dynamics model, but more precisely, real or actual values, especially measured values ​​or values ​​determined from measured values. In this regard, it is appropriate to use appropriate sensors, such as yaw rate sensors.

[0012] Then, the differences are calculated from the corresponding model-based values ​​and vehicle-based values—that is, from the values ​​with the same input variable. This results in a difference between the model-based values ​​and the vehicle-based values. The differences obtained in this way form the current data set. This current data set is then compared with a comparison data set to determine a consistency scale. If the consistency scale meets a pre-defined consistency criterion, the vehicle dynamics model is determined to be valid. Otherwise, the vehicle dynamics model is appropriately determined to be invalid.

[0013] A key advantage of the proposed method is its ability to be implemented in real time or during driving, specifically, by calculating the differences between multiple values ​​of the input parameters, preferably while the vehicle is in motion. This allows for continuous checking of whether a particular vehicle dynamics model can or cannot be used at any given time – the proposed method can also be implemented for multiple different vehicle dynamics models used in the vehicle.

[0014] As a comparison data set, it is preferable to use a data set that was obtained in the same manner as the current data set, but for example, within the scope of test measurements or during the first run of the vehicle, or only in other comparison time periods in which the vehicle dynamic model is considered or known to be sufficiently accurate in depicting reality.

[0015] Preferably, the differences between multiple values ​​of the input parameters are calculated sequentially, especially at predetermined time intervals. This ensures that the current data set is always up-to-date. Particularly suitable—and to limit the computational power required—is to determine the current data set only from a predetermined number of differences, that is, from a specific number of the last calculated differences.

[0016] It should also be mentioned that the values ​​of the input parameters are derived specifically from the actual behavior or operation of the vehicle and therefore can be repeated. When the steering angle is used as an input parameter, the values ​​should remain constant, or at least remain constant, when driving straight.

[0017] The comparison between the current data record and the comparison data record can be performed in such a way that not only the current data record but also the comparison data records—each containing a specific number of differences—are used as histograms, probability distributions, or frequency distributions, or are converted into or displayed as histograms, probability distributions, or frequency distributions. Suitable comparison methods exist for this purpose.

[0018] As a comparative method, the use of histogram differences, especially the so-called Jensen-Shannon divergence, is given preference. The histogram difference is a positive number generated by the difference between two histograms. It serves as a measure of the difference between the two histograms and can also be used as a measure of consistency—i.e., a consistency scale—or from which such a consistency scale can be derived. In particular, the Jensen-Shannon divergence represents a variation or modification of the so-called Kullback-Leibler divergence and will be explained in detail within the scope of the accompanying drawings.

[0019] As already mentioned, preferably, the vehicle dynamics model is determined to be invalid if the consistency metric does not meet a pre-defined consistency standard. In the case of Jensen-Shannon divergence, this means, for example, that the value obtained by the comparison method is greater than a specific comparison value or threshold. Correspondingly, when the value is less than (or equal to) the comparison value or threshold, the model is considered valid.

[0020] Especially when the model is determined to be invalid, a fault response can be introduced. Such a fault response can include, for example, introducing a safety response by using the vehicle dynamics model for a specific vehicle function, such as a driver assistance function or other function—for which the vehicle dynamics model is used—e.g., by introducing a restricted operating mode. However, as a fault response, fault storage recording, warnings to the vehicle driver (e.g., as a warning prompt on a display device), or switching to another driver assistance function or similar function can also be considered. Needless to say, such a fault response can be selected based on the type of vehicle dynamics model and / or its use. Multiple of the mentioned and / or other fault responses can also be introduced.

[0021] Overall, the proposed method improves vehicle safety, especially with the help of driver assistance systems and particularly during autonomous or automated driving.

[0022] The computing unit according to the invention, such as the controller of a motor vehicle, is particularly configured in terms of programming technology to implement the method according to the invention.

[0023] It is also advantageous to implement the method according to the invention in the form of a computer program or computer program product having program code for implementing all method steps, because this results in particularly low costs, especially if the controller used for execution is also used for other tasks and therefore already exists. Suitable data carriers for providing the computer program are, in particular, magnetic, optical, and electrical storage devices, such as hard disks, flash drives, EEPROMs, DVDs, etc. The program can also be downloaded via computer networks (Internet, intranet, etc.).

[0024] Other advantages and design solutions of the present invention are derived from the specification and drawings. Attached Figure Description

[0025] The present invention is schematically illustrated in the accompanying drawings by means of an embodiment and is described below with reference to the drawings.

[0026] Figure 1 A vehicle is schematically shown in which the method according to the invention can be implemented.

[0027] Figure 2The flowchart of a preferred embodiment of the method according to the invention is illustrated schematically.

[0028] Figures 3 to 5 It shows the method for in Figure 2 The diagram illustrates the method shown. Detailed Implementation

[0029] exist Figure 1 The diagram schematically illustrates a vehicle 100 in which the method according to the invention can be implemented. A computing unit 150 is exemplary disposed in the vehicle 100, on which a driver assistance system taking into account, for example, yaw rate can be run.

[0030] Furthermore, a yaw rate sensor 155 is provided for this purpose, by means of which the yaw rate can be measured and transmitted to the calculation unit 150. Additionally, a steering wheel 140 is shown, by means of which the steering angle can be adjusted. The steering angle can be detected, for example, by means of the steering angle sensor 145 and transmitted to the calculation unit 150. Needless to say, the steering angle can also be automatically adjusted or changed within the scope of driver assistance functions and / or autonomous or automated driving.

[0031] exist Figure 2 The diagram schematically illustrates a preferred embodiment of the method according to the invention. This method, for example, can be implemented in... Figure 1 The calculation is implemented on the shown computing unit 150. Figures 3 to 5 The diagram shows the method for using in Figure 2 The diagram illustrates the method shown.

[0032] Therefore, the values ​​W are repeatedly or sequentially detected or obtained for one or more input parameters E. E For example, the input parameter can be the steering angle, which can be as shown in the reference... Figure 1 To obtain it as explained.

[0033] Then, at the current value W of the steering angle E Based on this, the corresponding vehicle-based numerical value W of the output parameter A is obtained on the vehicle. F The output parameter A can be, for example, the deflection rate, which, or its value, can be as shown in the reference... Figure 1 As explained, it is acquired or measured using a deflection rate sensor.

[0034] In parallel with this, the current value W of the steering angle... E Based on this, the output parameter A, namely the deflection rate, is obtained or calculated using the vehicle dynamic model M.

[0035] This is used to calculate the yaw rate based on the steering angle δ. The vehicle dynamics model can be represented, for example, by the following differential equation:

[0036]

[0037] Here, β represents the sideslip angle, δ represents the steering angle, i represents the ratio between the steering angle (or steering wheel angle) and the wheel angle, v represents the vehicle speed, Θ represents the yaw inertia, and c f and c r Indicates the front and rear stiffness of the tire, l f and l r Let i represent the front and rear distances between the wheels and the center of gravity, and m represent the vehicle mass. Here, the wheel steering angle is therefore derived as the quotient of the proportion i and the steering angle δ.

[0038] The values ​​used for these parameters—apart from the input parameter δ and the vehicle speed v—are vehicle-specific parameters and are usually known or can be measured or calculated.

[0039] Therefore, in Figure 3 The graph above shows an exemplary curve of the steering angle as an input parameter E, expressed in seconds as time t. The graph below shows the corresponding curve V1 for the actual or measured evolution of the yaw rate as an output parameter A, and curve V2 for the evolution of the yaw rate obtained using a vehicle dynamics model. The numerical value W... F and W M It can be repeatedly acquired, for example, at intervals of one second or 100ms.

[0040] In addition, each is represented by a pair of corresponding numerical values ​​W. F and W M To obtain or form the difference W D Then, this difference W D The data is sent to a cache memory B, which stores all the differences obtained in this manner (in sequence over time).

[0041] Then, the differences W existing in buffer memory B D Forming the current data group H A Therefore, for example, it is possible to use a specific number, such as 300 of the most recent or last obtained differences, and store them in a buffer memory. It is also possible to consider storing only this number of values ​​in the buffer memory, while older values ​​are deleted or overwritten.

[0042] Then, this current dataset H is processed using a histogram, i.e., a frequency distribution. A Within the scope of the comparison method or comparison step 210, it is possible to compare the current data set H. A The comparison data set or reference data set H, also shown in histogram form. R The comparison is then performed. Here, the consistency scale ΔH is calculated and then compared with a pre-given threshold ΔH. S Compare them.

[0043] exist Figure 4 In the middle, the comparison data group H is shown on the left side in the form of a histogram. R And the current data group H is shown on the right. A Here, regarding 10 respectively -2 The difference W in units D The number of differences, N, is plotted. For comparison, both histograms must be normalized if necessary.

[0044] The comparison method will now be briefly and illustratively explained based on the Jensen-Shannon divergence already mentioned. The Jensen-Shannon divergence is based on the so-called Kullback-Leibler divergence. The Kullback-Leibler divergence measures how much the probability distribution P(x) differs from a second probability distribution Q(x). Here, the Kullback-Leibler divergence for discrete probability distributions on the same probability space is defined as follows:

[0045]

[0046] The Kullback-Leibler divergence is always non-negative, meaning D KL (P||Q)≥0, but is neither infinite nor symmetric. Therefore, in practice, the sum of two Kullback-Leibler divergences is often used.

[0047] D KL2 (P||Q)=D KL2 (Q||P)=D KL (P||Q)+D KL (P||Q)

[0048] This is used to ensure symmetry. However, in practical implementations, the infinite value of the Kullback-Leibler divergence causes problems. For this reason, the so-called Jensen-Shannon divergence D... JS It is preferred because it is symmetric and constrained and is based on the Kullback-Leibler divergence as follows:

[0049]

[0050] Where M = 0, 5·(P+Q) applies. Current data set H A And compare data group H R In the aforementioned notation, P and Q represent histograms or frequency distributions, and their comparison results in a consistency scale (or divergence scale) ΔH, which is positive in terms of Jensen-Shannon divergence.

[0051] For the output parameter in Figure 3 As shown in the graph below, the change curves yielded the result that, Figure 5 The curve showing the variation of such a consistency scale (or divergence scale) ΔH with respect to time t in seconds is presented. It can be clearly seen here that the consistency scale ΔH increases sharply, especially within the range of unusual and intense changes in the turning angle between 70 and 80 seconds.

[0052] In the current situation, the threshold ΔH S For example, a value of 0.25 can be selected, wherein if the consistency scale is below a threshold, i.e., if applicable: ΔH < ΔHs, then the vehicle dynamics model is determined to be valid.

[0053] In step 220, if the vehicle dynamics model is neither determined to be valid nor invalid, a fault response can be introduced as explained in detail above. However, if the vehicle dynamics model is determined to be valid, it can be used for the desired function as usual.

[0054] According to Figure 5 Therefore, it is possible that the vehicle dynamics model may be temporarily determined to be invalid and not used during this period. However, it may later become usable again.

Claims

1. Method for checking a vehicle dynamics model (M) of a vehicle (100), which check serves to determine whether the vehicle dynamics model (M) depicts the actual behavior of the vehicle sufficiently accurately and thereby can or is allowed to be used, wherein with the method a value of an output variable (A) can be determined from values of variable input variables (E), wherein a plurality of values (W E ) of the input quantity (E) are evaluated by means of the vehicle dynamic model (M) to a respectively belonging model-based value (W M ) of the output quantity (A), wherein a plurality of values (W E ) of the input quantity (E) are evaluated on the vehicle for a respective vehicle-based value (W F ) of the output quantity (A), wherein The vehicle-based numerical value (W F ) is a measured value, wherein a difference (W D ) is calculated from each of the model-based values (W M ) and the vehicle-based values (W F ) corresponding to each other, The difference (W) is obtained by means of the comparison method (210). D The current data group (H) A ) and comparison data group (H R The comparisons are performed on the data sets (H) and a consistency scale (ΔH) is calculated therein. R ) and the current data group (H) A Obtained in the same way, and wherein the vehicle dynamic model (M) is determined to be valid, if the consistency measure (AH) fulfills a pre-given consistency criterion (AH S ). wherein the input variables (E) are steering angles and the output variable (A) is a yaw rate.

2. The method according to claim 1, wherein the difference (W is determined during operation of the vehicle (100). D ).

3. The method according to claim 1 or 2, wherein the difference (W is repeatedly taken in time succession. D ).

4. The method according to claim 1 or 2, wherein the difference (W is repeatedly determined at predetermined time intervals. D ).

5. The method according to claim 3, wherein the current data set (H A ) is determined only from a predetermined number of difference values (W D ).​​​ 6. The method according to claim 1 or 2, wherein for a plurality of values (W E ) of the input quantity (E) vehicle-based values (W F ) belonging to the output quantity (A) are acquired on the vehicle by means of sensors (155).

7. Method according to claim 1 or 2, wherein as comparison method (210) a histogram difference is used.

8. Method according to claim 1 or 2, wherein as comparison method (210) a Jensen-Shannon divergence is used.

9. The method of claim 1 or 2, wherein, If the consistency measure (AH) does not satisfy a pre-defined consistency criterion (AH S ), the vehicle dynamic model (M) is determined to be invalid.

10. The method of claim 9, wherein, Then a fault reaction (220) is introduced.

11. The method of claim 1 or 2, wherein, If the vehicle dynamics model (M) is determined to be valid, the vehicle dynamics model (M) is used for a vehicle function.

12. Computing unit (150), which is set up to carry out all method steps of the method according to any one of claims 1 to 11.

13. Computer program product, which, when it is executed on a computing unit (150), causes the computing unit (150) to carry out all method steps of the method according to any one of claims 1 to 11.

14. Machine-readable storage medium, which has saved thereon a computer program, which is set up to carry out all method steps of the method according to any one of claims 1 to 11.

Images