Comparison of digital representations of driving conditions of vehicles

Abstract

Method (100) for comparing two digital representations (1, 2) of driving situations of a vehicle (50), wherein the digital representations (1, 2) each contain occupancy information (1a, 2a) about the occupancy of an environment (51) of the vehicle (50) by traffic-related objects, the method having the following steps: • dividing (110) an area (52) of the environment (51) of the vehicle (50) into a grid of partial areas (53); • determining (120) an occupancy (53a) of the partial areas (53) by traffic-related objects from the occupancy information (1a) in the first digital representation (1) and merging (130) the occupancy into a first fingerprint (1b) of a driving situation; • determining (140) an occupancy (53a') of the partial areas (53) by traffic-related objects from the occupancy information (2a) in the second digital representation (2) and merging (150) the occupancy into a second fingerprint (2b) of a driving situation; • determining (160) a similarity measure (4) between the first fingerprint (1b) and the second fingerprint (2b) according to a pre-given metric rule (3); • determining (180) that the two driving situations are identical or at least similar in response to the similarity measure (4) satisfying (170) a pre-given criterion (5).

Description

TECHNICAL FIELD

[0001] The invention relates to a comparison of digital representations of driving situations of a vehicle, which can be used, for example, to verify a simulation of a drive by means of a recording of an actually performed drive. BACKGROUND

[0002] A vehicle must prove itself to be traffic-safe in order to obtain an operating permit for road traffic. Therefore, all components and component groups that can influence traffic safety are detected in accordance with official requirements.

[0003] A system in which a vehicle can drive at least partially automatically in traffic must also prove itself explicitly to be safe in operation in any situation. This proof is very complex. From SAE automation level 3 onwards, it is possible that only 20% of the total expenditure in the implementation of new functions flows into the actual development, while the remaining 80% is required for the assurance.

[0004] A device is known from DE 10 2018 209 108 A1, which predicts the probability of an unexpected event in a technical system by means of neural network prediction technology.

[0005] In order to investigate the error behavior of at least partially automated vehicles that is not caused by a specific electrical or mechanical fault in the context of "Safety of the intended function" (SOTIF), the driving of these vehicles is simulated. Here, it is required to prove (verify) that the simulation is sufficiently close to reality. SUMMARY

[0006] Within the scope of the invention, a method for comparing two digital representations of driving situations of a vehicle is developed. These digital representations can each be present in arbitrary form. However, they each contain at least occupancy information about the occupancy of the environment of the vehicle by traffic-relevant objects.

[0007] Examples of traffic-relevant objects are, for example, driving road markings, driving road boundaries, other traffic participants, traffic signs and obstacles. The occupancy information can relate to all objects present in the environment of the vehicle or can relate only to specific types of objects. Thus, for example, occupancy information about driving road markings can be implemented separately from occupancy information about other traffic participants.

[0008] Within the scope of the application, a region of the vehicle's environment is divided into a grid of partial regions. This region does not necessarily extend around the entire vehicle. For example, if a comparison of driving situations is intended in terms of a particular driving task, then a region is sufficient which characterizes the driving situation in terms of the driving task. For example, if a vehicle which drives at least partially automatically is to be assessed as to how well it keeps to the lane, then as a region, the area of the driving road which lies ahead in the driving direction and which contains at least one lane boundary or other indicators of the lane in which the vehicle is currently driving is sufficient.

[0009] The occupation of the partial regions by the traffic-relevant objects is determined from the occupation information in the first digital representation and is merged into a fingerprint of the first driving situation. Likewise, the occupation of the partial regions by the traffic-relevant objects is determined from the occupation information in the second digital representation and is merged into a fingerprint of the second driving situation. A similarity measure between the first fingerprint and the second fingerprint is determined in accordance with a predefined metric rule. In response to the similarity measure value meeting a predefined criterion, it is determined that the two driving situations are identical or at least similar.

[0010] For example, the similarity measure can in particular be an arbitrary difference measure and / or distance measure which attributes a scalar or vectorial difference or distance to two fingerprints. Here, the difference measure and / or distance measure can also additionally be coupled to the corresponding specific application case. For example, if it is to be assessed in the driving situation how well an automated system recognizes or, respectively, keeps to the lane, then the elements of the fingerprint which characterize the lane can be weighted, for example, by means of elements of a weight matrix which is determined from the actual lane and / or the desired lane.

[0011] It has been recognized that in this way, a predefined first driving situation can in particular be compared very quickly with a plurality of other driving situations from a predefined reserve in order to re-identify the first driving situation in the reserve. For example, this can be used to call up additional information for the corresponding driving situation, which are saved in the reserve in connection with the driving situation.

[0012] For example, the stored driving situation reserve can comprise records of test drives completed by automated systems or also by human test drivers. If, for example, a driving situation is recognized again in such a reserve, a reaction that has already helped to successfully cope with this driving situation in an earlier test drive can be recalled, for example, and used in a similar manner to cope with the current situation. One exemplary application for this is a driving assistance system in a vehicle controlled by a human driver. If, for example, a driving situation is recognized in the operation of such a vehicle in which there is a threat of losing control and braking makes sense, this recommendation can be signaled to the driver. Alternatively, or also in combination with this, pressure is built up in the brake system, for example, in advance, in order to provide maximum braking force when the driver actually prompts braking.

[0013] The division of the area of the environment into partial areas in a grid, in combination with an analysis of the occupancy information in particular, can be responsible for both an abstraction and a compression of the information in the fingerprints. This is particularly advantageous in order to merge digital representations detected in completely different ways on a common basis and thus make them comparable. Thus, for example, a first representation can be known from a camera image, while a second representation is known from laser radar data. Likewise, for example, a first representation can be generated by simulation and a second representation by measurement. Further, the problem is eliminated that a sensor-based detection of a driving situation can essentially only be reproduced to a limited extent, which makes re-identification difficult. Thus, for example, two camera images directly following one another, which are completed from the same perspective, of the same driving situation, are far from identical on the plane of the recorded pixel values. Rather, what objects are present where does not change, and the division into a grid of partial areas also absorbs fluctuations related thereto to a certain extent.

[0014] In a particularly advantageous configuration, it is detected in binary in one bit each whether a partial area is occupied by at least one traffic-relevant object. The conclusion can relate to all considered types of traffic-relevant objects or also only to specific types of objects. The comparison of bits and / or bit sequences can be carried out in a machine-like manner particularly efficiently. Thus, for example, a search for a specific driving situation in a very large driving situation reserve can also be carried out quickly.

[0015] Thus, for example, in a particularly advantageous configuration, a first digital representation is derived from at least one simulation of a driving situation, and a second digital representation is derived from a record of measurement data recorded during a drive of a vehicle by means of at least one sensor. The measurement data can be, for example, unclassified time courses having a volume in the order of magnitude of 10000 driving hours There is. Here, too, the transition between the driving situations that follow one another is not annotated. With the help of the binary-coded fingerprints, it is also possible to quickly search the thus large representation reserves.

[0016] It is particularly advantageous to merge the bits detected for all partial areas into one binary number as a fingerprint. For example, it is possible to check two such binary numbers for exact identity in a unique machine command. It is therefore also advantageous to divide the area of the environment into a number of partial areas that is a power of two. Advantageously, this power of two then corresponds to the width of a register on the hardware platform used for the comparison, i.e. 64 bits on a 64-bit system.

[0017] It is particularly advantageous if the significance of a bit in the binary number is the smaller, or the larger, the more important the object in the partial area corresponding thereto is for the behavior planning of the vehicle in the driving situation. This, for example, not only simplifies the search for exact identity of the binary numbers, but also simplifies deviations of the binary numbers that are permitted within a given tolerance limit. If the most important features of the driving situation are the same, then the bits of the least significant value or the most significant value of the binary number are also the same, respectively.

[0018] Thus, for example, the similarity measure can depend on the length of the identical bit sequence in the two fingerprints. This length can be determined particularly efficiently and then directly provide a conclusion as to how large the deviation in quantity between the two binary numbers can be at most.

[0019] For example, if it is to be investigated how well a vehicle that is at least partially automated drives, then the lane boundary of the area in front of the vehicle can be observed in particular. If the lane boundary is at the position expected of it in the lane that is expected to be driven well, then this is less relevant for the behavior planning, since the current behavior is good in the sense of the driving task and does not have to change suddenly. The further the lane boundary is from the expected position, the further the vehicle is outside its target lane, the more important the countersteering is. Here, for example, the viewing angle of the observation can additionally also be taken into account. If the lane boundary is laterally offset on the horizon at a certain distance in front, then this corresponds to a significantly smaller angular deviation of the course of the vehicle than if the lane boundary directly in front of the vehicle is laterally offset by the same amount.

[0020] In another advantageous configuration, the similarity measure depends on the Hamming distance between the two fingerprints. The Hamming distance measures how many bits differ from one another between the two fingerprints. This is particularly advantageous when all partial areas of the area of the environment of the vehicle are essentially equally important for the behavior planning of the vehicle.

[0021] In another particularly advantageous configuration, the detected occupancy for all partial areas is respectively merged into a matrix and a fingerprint is formed therefrom. In this way, for example, the two-dimensional structure of the spatial area viewed as a whole is also retained in the fingerprint. Additional knowledge about the two-dimensional structure can thus also be used for possible pre-processing of the matrix into a fingerprint.

[0022] Thus, for example, at least the matrix resulting from one of the representations can be converted into a fingerprint in the case of filtering using at least one filter kernel. By selecting a suitable filter kernel, it is possible to tailor the similarity measure to specific aspects of the driving situation which are important for the comparison. For example, it is thus possible to achieve that the similarity measure measures exclusively to what extent the lanes followed according to the two driving situations under investigation are identical.

[0023] To this end, the similarity measure may, for example, comprise the mean or median of the elements in the element-wise product of the two fingerprints.

[0024] The comparison of the digital representations of two driving situations may, for example, be used to verify a simulation of a vehicle's driving. In this connection, verification means, inter alia, that the simulation is "anchored" in reality: if it is possible to find a known reference driving situation for a sufficient number of simulated driving situations which further develops exactly the same as the respective simulated situation, then reality is reproduced with high probability by the simulation with sufficient accuracy.

[0025] The invention therefore also provides a method for verifying a simulation of a vehicle's driving.

[0026] The method begins with the conversion of at least one simulated driving situation of a simulated driving into a digital representation. Further, digital representations of a plurality of reference driving situations within a reference driving are taken from the time course of the reference driving.

[0027] The digital representations of the simulated driving situations are compared with the digital representations of the reference driving situations in accordance with the method described previously. From the results of these comparisons, at least one reference driving situation within the reference driving which is identical or at least similar to the simulated driving situation is determined. What is to be regarded as identical or similar is determined by the similarity measure used in the search.

[0028] The course of the following of the simulated journey is compared with the course of the following of the reference journey sought in the reference journey. This comparison can be made, for example, with reference to the respective complete courses. But the comparison can also assess, for example, whether the two courses end up in the same or a similar result. It is thus less important, for example, whether the lane change is made quickly and with a small curve radius or slowly and with a larger curve radius. What is important is only whether the lane change is made correctly at the end.

[0029] In response to the course of the following of the simulated journey agreeing with the course of the following of the reference journey according to predefined criteria, it is determined that the simulation of the journey is at least valid for the journey situation investigated.

[0030] It is not important here whether the course of the following of the simulated journey and / or the course of the following of the reference journey is objectively correct in the respective case. Rather, the occurrence of the same type of error in the simulated journey and in the reference journey is a strong indication in the direction that the simulation sufficiently accurately reproduces reality.

[0031] An example of this is a vehicle in at least partially automated journey autonomously turning off at a motorway exit. The specific features of such an exit, such as in particular the new road state, can induce the behavior planner of the vehicle to assume that not the right lane of the current journey, but the exit is the continuous journey road. If this error occurs not only in the simulation, but also in the same or similar reference situation during one or more reference journeys, the simulation with high probability covers the mechanisms of action that lead to the occurrence of the error. As a result, the simulation can then be used to make the causes of the error findable, so that these causes can subsequently be eliminated.

[0032] If a simulation model is available that has been verified in the manner described above, such a simulation model can be used to automatically generate a large number of variations of the journey situation. The error causes can then be inferred from the corresponding progress of the variations in the simulation. Thus, for example, a pattern can be derived that spontaneous journey deviations from the motorway tend to occur when the motorway cover is new, it is raining and the temperature is below 15°C.

[0033] In a particularly advantageous configuration, indicators of the achieved braking danger and / or steering potential required for danger mitigation are used for the comparison of the courses of the following of the simulated journey on the one hand and the reference journey on the other hand. These are indicators for the driving maneuver Key indicators of the result achieved at the end.

[0034] The methods described previously can be computer-implemented, in particular, for example, and thus embodied in software. The present application also relates to a computer program which contains machine-readable instructions which, when the instructions are executed on one or more computers, cause the one or more computers to implement one of the described methods. In this sense, a controller for a vehicle and an embedded system for a technical device can also be regarded as computers, which are likewise able to implement machine-readable instructions.

[0035] Likewise, the present application also relates to a machine-readable data carrier and / or a download product with the computer program. A download product is a digital product which can be transferred by a data network, i.e. can be downloaded by a user of a data network, which can be sold, for example, in an online shop for immediate download.

[0036] Furthermore, a computer can be equipped with the computer program, with the machine-readable data carrier or with the download product. BRIEF DESCRIPTION OF DRAWINGS

[0037] Further measures for improving the application are represented in more detail in the following together with a description of preferred embodiments of the application according to the drawings.

[0038] Figure 1 An embodiment of a method 100 for comparing two numerical representations of driving situations of a vehicle is shown;

[0039] Figure 2 An embodiment of a method 200 for verifying a simulation of a driving of a vehicle is shown;

[0040] Figure 3 An exemplary division of a region of an environment of a vehicle into partial regions is shown;

[0041] Figure 4 An exemplary determination of a similarity of two driving situations in the case of a filtering using a filter kernel 6 in the formation of a fingerprint 1b is shown. DETAILED DESCRIPTION

[0042] Figure 1is a schematic flow chart of an embodiment of a method 100 for comparing two digital representations 1, 2 of driving situations of a vehicle 50. In step 105, a first digital representation 1 is derived from at least one simulation of a driving situation. In step 106, a second digital representation 2 is derived from at least one record of measurement data recorded by means of at least one sensor during a driving of the vehicle 50. Both digital representations 1, 2 comprise occupancy information 1a, 2a about an occupancy of an environment 51 of the vehicle 50 with traffic-related objects, respectively. In step 110, an area 52 of the environment 51 is divided into a grid 53 of partial areas. In step 120, an occupancy 53a of the partial areas 53 with traffic-related objects is derived from the occupancy information 1a. This can be done in particular in a binary form of bits according to block 121, which show for each respective partial area 53 whether it is occupied with an object or not.

[0043] In step 130, the occupancies 53a are merged into a fingerprint 1b of the first driving situation. If the occupancies 53a are bits, these can be merged into a binary number as the fingerprint 1b according to block 131. In general, the occupancies 53a can also be merged into a matrix M according to block 132, and the fingerprint 1b can be formed therefrom. In particular, the matrix M can be transformed into the fingerprint 1b with filtering using at least one filter kernel 6 according to block 132a.

[0044] In step 140, an occupancy 53a' of the partial areas 53 with traffic-related objects is derived from the occupancy information 2a, similar to step 120. This can be done in particular in a binary form of bits according to block 141, which show for each respective partial area 53 whether it is occupied with an object or not.

[0045] In step 150, the occupancies 53a' are merged into a fingerprint 2b of the first driving situation, similar to step 130. If the occupancies 53a' are bits, these can be merged into a binary number as the fingerprint 1b according to block 151. In general, the occupancies 53a can also be merged into a matrix M according to block 152, and the fingerprint 2b can be formed therefrom. In particular, the matrix M can be transformed into the fingerprint 2b with filtering using at least one filter kernel 6 according to block 152a.

[0046] In step 160, a similarity measure value 4 between the first fingerprint 1b and the second fingerprint 2b is derived according to a pre-given metric rule 3. In step 170, it is checked whether the similarity measure value meets a pre-given criterion 5. If this is the case (true value 1), it is determined in step 180 that the two driving situations characterized by the digital representations 1 and 2 are identical or at least similar.

[0047] Figure 2is a schematic flow chart of an embodiment of a method 200 for verifying a simulation of a journey of a vehicle 50.

[0048] In step 210, at least one simulated journey condition 1* of the simulated journey is converted into a digital representation 1. In step 220, a digital representation 2 of a plurality of reference journey conditions within the reference journey is derived from a time course 2z of the reference journey. In step 230, the digital representation 1 of the simulated journey condition is compared with the digital representation 2 of the reference journey condition according to the previously described method 100.

[0049] From the similarity 4 derived in these comparisons 230, in step 240 at least one reference journey condition 2* within the reference journey is derived. This can be, for example, the reference journey condition 2* for which the similarity 4 is greatest or is above a threshold value of the similarity measure.

[0050] In step 250, the course 1# of the simulated journey following the simulated journey condition 1* is compared with the course 2# of the reference journey following the derived reference journey condition 2*. For this, in particular, for example according to block 251, indicators of the braking danger reached in the course 1#, 2# of the simulated journey on the one hand and of the reference journey on the other hand and / or the steering potential required for danger mitigation can be used.

[0051] In step 260, the result 250a of the comparison 250 is checked in terms of whether the course 1# of the simulated journey following the simulated journey condition 1* is in agreement with the course 2# of the reference journey following the reference journey condition 2* according to a pre-given criterion 7. If this is the case (true value 1), it is determined in step 270 that the simulation of the journey is at least valid for the journey condition 1* under investigation.

[0052] Figure 3 It is exemplarily shown how a region 52 is selected from the environment 51 of the vehicle 50 and how this region is divided into partial regions 53. In this example, the region 52 is to the left and in front of the vehicle 50 in the direction of travel. The region 52 is 50 m long and 3.75 m wide. The region contains a lane boundary F.

[0053] The region 52 is divided into 63 partial regions 53. These partial regions 53 contain bit indices between 2 and 64 which state how important the occurrence of a part of the travel road boundary F in the corresponding partial region is for the behavior planning of the vehicle 50. The higher the bit index, the less important it is. Thus, it is less important if the lane boundary F remains at its nominal position. However, it is more important if the lane boundary F moves to the edge of the region 52, more specifically, the closer it is to the vehicle 50 and the greater the angular deviation of the curve of the vehicle 50 is thus, the more important it is. A bit with the bit index 1 is reserved to show an error upon detection.

[0054] The digital representation 1, 2 of the driving situation can thus be compressed to a unique 64-bit number as a fingerprint 1b, 2b.

[0055] Figure 4 It is schematically illustrated how the fingerprint 1b is calculated in the case of filtering using the filter kernel 6 and is then calculated together with the second fingerprint 2b to the similarity 4. In the case of the first driving situation 1, the filter kernel 6 is applied to the matrix M in which the occupancy 53a is incorporated. Figure 1 In the example shown, in the two driving situations characterized by the representations 1 and 2, the lane position is slightly offset in the horizontal direction. In the matrix, the point density in a single element represents the value of this element, respectively. The greater the point density in an element, the greater the value of this element.

[0056] The first representation 1 relates to a driving situation in which the lane position is exactly in the middle. The matrix M in which the occupancy 53a is incorporated thus has a column with particularly large values in the middle. By filtering the matrix M using the filter kernel 6, in the fingerprint 1b resulting therefrom, the high values are slightly "averaged" to the left and to the right.

[0057] The second representation 2 relates to a driving situation in which the lane position is slightly offset to the left. In the unfiltered fingerprint 2b in which the occupancy 53a' is incorporated, there are thus particularly large values in the column to the left in the middle. If the fingerprint 2b is multiplied element-wise with the fingerprint 1b, an intermediate result is produced in which, to the left of the middle, a column with elevated values is produced. The formation of the mean or median of all elements of this intermediate result provides the sought similarity 4. If, in the second driving situation, the lane position is likewise in the center, this similarity is clearly represented as being higher.

Claims

1. A method (100) for comparing two digital representations (1, 2) of a driving situation of a vehicle (50), wherein, The digital representations (1, 2) each contain occupancy information (1a, 2a) about the occupancy of the environment (51) of the vehicle (50) by traffic-related objects, the method having the following steps: • dividing (110) the area (52) of the environment (51) of the vehicle (50) into partial areas of a grid (53); • determining (120) the occupancy (53a) of the partial areas (53) by traffic-related objects from the occupancy information (1a) in the first digital representation (1) and merging (130) the occupancy into a first driving situation fingerprint (1b); • determining (140) the occupancy (53a') of the partial areas (53) by traffic-related objects from the occupancy information (2a) in the second digital representation (2) and merging (150) the occupancy into a second driving situation fingerprint (2b); • determining (160) a similarity measure (4) between the first driving situation fingerprint (1b) and the second driving situation fingerprint (2b) from a pre-given metric rule (3); • determining (180), in response to the similarity measure (4) satisfying (170) a pre-given criterion (5), that the two driving situations are identical or at least similar, wherein it is detected (121, 141) in one bit each whether a partial area (53) is occupied by at least one traffic-related object or not, wherein the bits detected for all partial areas (53) are merged (131, 151) into a binary number as a fingerprint (1b, 2b), wherein the smaller or larger the significant bit value of one bit in the binary number, the more important the behavior of the object in the partial area (53) corresponding to this bit for the planning of the vehicle (50) in the driving situation, wherein the similarity measure (4) depends on the length of the sequence of bits that is identical in both fingerprints (1b, 2b).

2. The method (100) of claim 1, wherein The similarity measure (4) depends on the Hamming distance between the two fingerprints (1b, 2b).

3. The method (100) of claim 1, wherein The occupancies (53a) detected for all partial areas are each merged into a matrix (M) and the fingerprint (1b, 2b) is formed (132, 152) therefrom.

4. The method (100) of claim 3, wherein The matrix (M) resulting from one of the representations (1, 2) is at least converted (132a, 152a) into a fingerprint (1b, 2b) using filtering with at least one filter kernel (6).

5. The method (100) according to any one of claims 3 to 4, wherein, The similarity measure (4) contains the mean or median of the elements in the element-wise product of the two fingerprints (1b, 2b).

6. The method (100) according to any one of claims 1 to 4, wherein The first digital representation (1) is determined (105) from at least one simulation of a driving situation, and wherein the second digital representation (2) is determined (106) from at least one recording of measurement data recorded by means of at least one sensor during a drive of the vehicle (50).

7. A method (200) for verifying a simulation of a drive of a vehicle (50), the method having the steps: • converting (210) at least one simulated driving situation (1*) of a simulated drive into a digital representation (1); • determining (220) a similarity measure (4) between the first driving situation fingerprint (1b) and the second driving situation fingerprint (2b) from a pre-given metric rule (3); • determining (230), in response to the similarity measure (4) satisfying (220) a pre-given criterion (5), that the two driving situations are identical or at least similar, wherein it is detected (121, 141) in one bit each whether a partial area (53) is occupied by at least one traffic-related object or not, wherein the bits detected for all partial areas (53) are merged (131, 151) into a binary number as a fingerprint (1b, 2b), wherein the smaller or larger the significant bit value of one bit in the binary number, the more important the behavior of the object in the partial area (53) corresponding to this bit for the planning of the vehicle (50) in the driving situation, wherein the similarity measure (4) depends on the length of the sequence of bits that is identical in both fingerprints (1b, 2b). The similarity measure (4) depends on the Hamming distance between the two fingerprints (1b, 2b). The occupancies (53a) detected for all partial areas are each merged into a matrix (M) and the fingerprint (1b, 2b) is formed (132, 152) therefrom. The matrix (M) resulting from one of the representations (1, 2) is at least converted (132a, 152a) into a fingerprint (1b, 2b) using filtering with at least one filter kernel (6). The similarity measure (4) contains the mean or median of the elements in the element-wise product of the two fingerprints (1b, 2b). The first digital representation (1) is determined (105) from at least one simulation of a driving situation, and wherein the second digital representation (2) is determined (106) from at least one recording of measurement data recorded by means of at least one sensor during a drive of the vehicle (50). • deriving (220) a digital representation (2) of a plurality of reference driving situations within the reference drive from a time history (2z) of the reference drive; • comparing (230) the digital representation (1) of the simulated driving situation with the digital representation (2) of the reference driving situation according to the method (100) of any one of claims 1 to 6; • deriving (240) at least one reference driving situation (2*) within the reference drive which is identical or at least similar to the simulated driving situation (1*) from the result (230a) of the comparison; • comparing (250) the course (1#) of the simulated driving from the simulated driving situation (1*) with the course (2#) of the reference driving from the derived reference driving situation (2*) of the reference drive; • determining (270) that the simulation of the drive is valid at least for the investigated driving situation (1*) in response to the course (1#) of the simulated driving being identical (260) to the course (2#) of the reference driving according to a predefined criterion (7).

8. The method (200) of claim 7, wherein, In the course (1#, 2#) of the simulated driving and in the course (1#, 2#) of the reference driving, respectively, an indicator of the braking danger reached and / or the steering potential required for danger mitigation is used (251) for the comparison (250) of these courses (1#, 2#).

9. A computer program comprising machine-readable instructions which, when executed on one or more computers, cause the one or more computers to carry out the method (100, 200) according to any one of claims 1 to 8.

10. A machine-readable data carrier and / or a download product having the computer program according to claim 9.

11. A computer having the computer program according to claim 9 and / or the machine-readable data carrier and / or a download product according to claim 10.

Images