Method for determining the criticality of the evasive behavior of an at least partially automated vehicle

Abstract

A general aspect of the present disclosure relates to a method for determining the criticality of the evasive behavior of an at least partially automated vehicle moving at an instantaneous speed. The method comprises calculating a plurality of evasive trajectories of the vehicle with respect to a referenced collision object by combining different longitudinal accelerations with different lateral accelerations. The method further comprises determining an optimal trajectory from the plurality of evasive trajectories, where the minimum distance between the vehicle and the collision object is maximized. The method further comprises determining a critical acceleration vector of an associated critical trajectory, where the minimum distance reaches a near-collision threshold.The procedure further includes determining a criticality of the optimal trajectory based on the critical acceleration vector and an optimal acceleration vector corresponding to the optimal trajectory.

Description

Technical area

[0001] The present disclosure relates to driver assistance systems and methods with at least partially automated driving functions. Specifically, a computer-implemented method for determining the criticality of the evasive behavior of an at least partially automated vehicle is disclosed. State of the art

[0002] Driver assistance systems are known for warning drivers and for steering or braking intervention or corresponding driver warnings in dangerous situations. In the further development of driver assistance systems into partially automated, automated, or highly automated driving systems, various approaches to collision avoidance or collision consequence mitigation are being pursued. In principle, such driving systems can predict and / or execute automated evasive maneuvers for collision avoidance, which include pure braking, pure steering, or combined braking and steering relative to a collision object detected, for example, by sensor data. By planning the vehicle's trajectory relative to the collision object, it can be determined whether an evasive maneuver is necessary and / or possible. Of great importance for such driving systems is the evaluation of evasive behavior with regard to safety.In particular, such driving systems must be safeguarded against risks (i.e., validated) before market launch. A safety assessment is also relevant for the optimal selection of an evasive maneuver with a minimal or at least reduced risk. Various criticality metrics are available for this purpose, enabling a quantitative evaluation of the driving behavior or evasive maneuvers. A fundamental distinction can be made between subcritical cases, in which a collision can be avoided, and critical cases, in which a collision is unavoidable. In particular, criticality metrics should reliably evaluate driving situations at both low and high speeds.

[0003] The published patent application DE 10 2007 007 640 A1 describes a method for detecting critical accident situations in road traffic. For this purpose, a virtual lane is calculated based on a few reference points in the drivable space around the vehicle. This virtual lane is used to check whether and when other road users cross the virtual lane and what acceleration would be necessary to avoid a collision. The situation is classified as critical if, after optimizing the virtual lane, the acceleration required to avoid a collision exceeds a threshold.

[0004] Published patent application DE 10 2015 221 612 A1 describes a method for calculating a vehicle trajectory. It generates a series of possible trajectories with different end times and target states, with the acceleration curve of each trajectory limited by a specified maximum acceleration. For each trajectory, an evaluation function is calculated to select the most favorable one. This is then determined as the trajectory to be driven after checking whether it complies with the vehicle's dynamic limits.

[0005] The publication Cheng Wang et al.: Acceleration-based collision criticality metric for holistic online safety assessment in automated driving. IEEE Access, 10, June 27, 2022, 70662-70674, proposes a new acceleration-based metric for collision criticality assessment, applicable to both longitudinal and lateral directions. This metric calculates the minimum required acceleration or deceleration for various driving maneuvers (braking, swerving) to avoid a collision and combines them into an overall value. The metric has been validated in simulations and with a naturalistic driving dataset, demonstrating higher accuracy and recall compared to the time-to-collision (TTC).

[0006] Furthermore, the metric combines the strengths of TTC and time gap (THW) to evaluate critical situations in the longitudinal area.

[0007] The present disclosure addresses the problem of how a quantitative assessment of an evasive behavior of an at least partially automated vehicle can be provided more reliably. Disclosure of the invention

[0008] A first general aspect of the present disclosure relates to a computer-implemented method for determining a criticality of the evasive behavior of an at least partially automated vehicle, wherein the at least partially automated vehicle is moving at a current speed. The method comprises calculating a plurality of evasive trajectories of the vehicle with respect to a referenced collision object by combining various longitudinal accelerations with various lateral accelerations. The method further comprises determining an optimal trajectory from the plurality of evasive trajectories in which a minimum distance between the vehicle and the collision object is maximum. The method further comprises determining a critical acceleration vector of an associated critical trajectory in which the minimum distance reaches a near-collision threshold.The method further comprises determining a criticality of the optimal trajectory based on the critical acceleration vector and an optimal acceleration vector associated with the optimal trajectory. In embodiments, determining the critical acceleration vector may comprise reducing the magnitude of the optimal acceleration vector down to the critical acceleration vector at which the minimum distance reaches the threshold of a near-collision. In embodiments, the criticality of the optimal trajectory may be determined by forming the ratio of the magnitude of the critical acceleration vector to the magnitude of the optimal acceleration vector.A second general aspect of the present disclosure relates to a computer system configured to execute the computer-implemented method for determining a criticality of the evasive behavior of an at least partially automated vehicle according to the first general aspect (or an embodiment thereof).

[0009] A third general aspect of the present disclosure relates to a computer program comprising instructions which, when the computer program is executed by a computer system, cause the computer system to execute the computer-implemented method for determining a criticality of the evasive behavior of an at least partially automated vehicle according to the first general aspect (or an embodiment thereof).

[0010] A fourth general aspect of the present disclosure relates to a computer-readable medium or signal storing and / or containing the computer program according to the third general aspect (or an embodiment thereof).

[0011] A fifth general aspect of the present disclosure relates to a method for testing and / or validating an at least partially automated driving function system. In the method, the criticality determined by the computer-implemented method according to the first general aspect (or an embodiment thereof) is used as an evaluation criterion for the release of the at least partially automated driving function system. A sixth general aspect of the present disclosure relates to a method for setting up an at least partially automated driving function system. In the method, the setting up of at least partially automated driving functions, in particular the setting up of rules for evasive maneuvers, is carried out at least based on the criticality determined by the computer-implemented method according to the first general aspect (or an embodiment thereof).

[0012] The computer-implemented method proposed in this disclosure according to the first general aspect (or an embodiment thereof) can be used to obtain a quantitative assessment of a driving behavior, in particular an evasive behavior, of an at least partially automated vehicle. In particular, the computer-implemented method can be used to evaluate various evasive maneuvers with regard to a maximum usable vehicle capability. In other words, the proposed determination of criticality can be used to assess which proportion of the total vehicle capability must be used to avoid a collision. This enables better comparability of the criticalities in different driving situations, in particular driving situations with different instantaneous speeds.In other words, the determined criticalities are more comparable over large speed ranges than the minimum distances. This can improve the selection and validation of evasive maneuvers with regard to a collision risk and / or with regard to mitigating the consequences of a collision. Accordingly, the proposed method can improve a partially automated driving function. A further advantage of the proposed computer-implemented method can be seen in the fact that the quantitative recognition of the utilized proportion of the overall vehicle capability allows the intensity (e.g., braking intensity and / or steering intensity or longitudinal acceleration and / or lateral acceleration) of a driving maneuver, for example an evasive maneuver, to be reduced or increased. This, in turn, can result in advantages in terms of user satisfaction or wear reduction.

[0013] Some terms are used in this disclosure as follows: A “longitudinal acceleration” can be regarded as the amount of acceleration in the longitudinal direction, i.e. in the longitudinal direction of the at least partially automated vehicle under consideration.

[0014] A “lateral acceleration” can be regarded as the amount of acceleration in the lateral direction, i.e. in the sideways direction of the at least partially automated vehicle under consideration.

[0015] An "acceleration vector a→'' is the resultant of longitudinal acceleration and lateral acceleration.

[0016] A "vehicle" can be any device that transports passengers and / or cargo. A vehicle can be a motor vehicle (e.g., a car or a truck), but also a rail vehicle. A vehicle can also be a motorized two- or three-wheeler. However, floating and flying devices can also be vehicles. Vehicles can be at least partially autonomous or assisted.

[0017] The term “at least partially automated” includes devices and / or processes that are partially automated, automated and / or highly automated. Short description of the characters Fig. 1 schematically illustrates a method for determining a criticality of the evasive behavior of an at least partially automated vehicle, as well as the use of the determined criticality: - in a method for testing and / or validating an at least partially automated driving function system, and - in a method for setting up an at least partially automated driving function system. Fig. Figure 2A schematically illustrates the determination of an exemplary avoidance trajectory by combining a longitudinal acceleration with a lateral acceleration in a maximum friction force model. Fig. Figure 2B schematically illustrates the determination of an exemplary avoidance trajectory by combining a longitudinal acceleration with a lateral acceleration in a maximum friction force model. Fig. 3 schematically illustrates an exemplary evasive maneuver of an at least partially automated vehicle that can avoid a collision object at a distance dmin according to an exemplary evasive trajectory. Detailed description

[0018] First, with regard to the Fig. 1, Fig. 2A and Fig. 2B discusses the techniques of the present disclosure. With respect to Fig. 3, possible results and advantages resulting from the computer-implemented method 100 disclosed herein for determining a criticality KRIT of the evasive behavior of an at least partially automated vehicle 1 are discussed.

[0019] Fig. 1 is a flow chart illustrating possible steps of the computer-implemented method 100 for determining a criticality KRIT of the evasive behavior of an at least partially automated vehicle 1, wherein the vehicle 1 is traveling at a current speed v→ The computer-implemented method 100 includes calculating 110 a plurality of avoidance trajectories τ iof the vehicle with respect to a referenced collision object 3 by combining different longitudinal accelerations 12 with different lateral accelerations 14, 14a. Furthermore, in step 120, an optimal trajectory τ optimal from the majority of avoidance trajectories τ i determined at which a minimum distance d min between the vehicle 1 and the collision object 3 is maximum. The method 100 further comprises determining 130 a critical acceleration vector a→crit a corresponding critical trajectory τ krit , which reaches a threshold of a near collision. The method further comprises determining a criticality CRIT optimal the optimal trajectory τ optimal based on the critical acceleration vector a→crit and one to the optimal trajectory τ optimal corresponding optimal acceleration vector a→optimal.

[0020] As continued in Fig. 1, the present disclosure also relates to a method 200 for testing and / or securing an at least partially automated driving function system 2. In the method 200, the criticality KRIT determined by the computer-implemented method 100 is used as an evaluation criterion for the release of the at least partially automated driving function system 2.

[0021] Further with regard to Fig. 1, the present disclosure also relates to a method 300 for setting up an at least partially automated driving function system 2. In the method 300, the setting up of at least partially automated driving functions is carried out at least based on the criticality KRIT determined by the computer-implemented method 100. In embodiments, the setting up of at least partially automated driving functions can include setting up rules for evasive maneuvers. In embodiments, a control unit of the at least partially automated driving function system 2 can be set up at least based on the criticality KRIT determined by the computer-implemented method 100. In embodiments, a back-end system can be set up that is designed to communicate with the at least partially automated driving function system 2.In embodiments of method 300, the setting up may include loading software into a memory of the at least partially automated driving function system 2. In embodiments, the software may be loaded over-the-air.

[0022] Based on the Fig. 2A and Fig. 2B, further embodiments of the computer-implemented method 100 are explained below. Fig. 2A and Fig. 2B show an exemplary methodology for determining the avoidance trajectories τ i or their corresponding acceleration vectors a→i. In detail, the Fig. 2A and Fig. 2B shows a maximum friction force model 10. This is a model that takes into account the maximum forces or accelerations that can be transmitted by a vehicle to its environment. In the present example, a Kamm circle model 10, familiar to road vehicle experts, is used. Accordingly, depending on vehicle and surface specifications, a limit range of maximum transmittable forces or accelerations can be schematically represented by the circle shown. The axis 12, 12a represents longitudinal accelerations 12, 12a. Starting from the origin upwards, the axis 12 can represent braking accelerations 12, decelerations, or reverse accelerations. Starting from the origin downwards, the axis 12a can represent driving accelerations 12a or forward accelerations. The axis 14, 14a represents lateral accelerations 14, 14a, which can be caused, for example, by steering the vehicle 1.Axis 14, extending from the origin to the right, can indicate lateral accelerations directed to the right (for example, by steering to the left). Axis 14a, extending from the origin to the left, can indicate lateral accelerations directed to the left (for example, by steering to the right). The representation of right-left accelerations and forward-backward acceleration explained here as an example can also be defined alternatively in other embodiments. Essential for the representation is the representation of longitudinal accelerations or forces orthogonal to lateral accelerations or forces.

[0023] The combining of different longitudinal accelerations 12 with different lateral accelerations 14 included in step 110 can be carried out according to the Fig. 2A, therefore, involves combining the maximum usable braking capacity with the maximum usable steering capacity. In other words, the maximum usable vehicle capability with respect to steering and / or braking (i.e., the maximum transferable friction force) is determined according to the maximum friction force model 10. The various longitudinal accelerations 12 can be based on different braking intensities. The various lateral accelerations 14 can be based on different steering intensities. Different ratios of longitudinal accelerations 12 and lateral accelerations 14 can be combined by selecting different points along the circle and the respective resulting amounts 12 i and 14 i This is in Fig. 2A is shown schematically by the curved arrow and the crosses on the circle. An example of the avoidance trajectory τ ithe longitudinal acceleration 12 i and lateral acceleration 14 i The resultant of longitudinal acceleration 12 i and lateral accelerations 14 i forms the corresponding avoidance trajectory τ i corresponding acceleration vector a→i. Using trajectory calculation models known to the expert, the current speed v→ of the vehicle the trajectory or avoidance trajectory τ i Alternatively, the majority of avoidance trajectories τ i based on the current speed v→ by combining different braking intensities with different steering intensities. The calculation of the majority of avoidance trajectories τ i can be calculated from, for example, 2 to 1000, 10 to 500 or 50 to 100 avoidance trajectories τ iIn some embodiments, more than 1000 avoidance trajectories τ i Even if the Fig. 2A, only one side of lateral acceleration 14 is shown, in embodiments, lateral accelerations 14a to the other side can also be taken into account. By combining different longitudinal accelerations 12 with different lateral accelerations 14, pure braking, pure steering, and values ​​in between can be taken into account. Alternatively, by combining different longitudinal accelerations 12 with different lateral accelerations 14 according to the maximum friction force model 10, different maximum transferable acceleration vectors can be determined. a→i. from a braking vector 12 i and a steering vector 14 i be formed.

[0024] As explained in detail below in the Fig. 3, for each calculated avoidance trajectory τ i the minimum distance d min between the vehicle 1 and the collision object 3. In embodiments, the respective minimum distances d min the majority of avoidance trajectories τ i can be compared with each other. The avoidance trajectory τ i the majority of avoidance trajectories τ i with the largest minimum distance d min as optimal avoidance trajectory τ optimal selected or determined.

[0025] In Fig. 2B is an example optimal avoidance trajectory τ optimal with corresponding optimal acceleration vector a→optimal and corresponding optimal longitudinal acceleration 12 opt and optimal lateral acceleration 14 opt As further shown in Fig. 2B, determining the critical acceleration vector a→crit reducing the magnitude of the optimal acceleration vector a→optimal up to the critical acceleration vector a→crit where the minimum distance d min reaches the limit of a near collision. In embodiments, as in Fig. 2B, for example, by the arrow along the acceleration vector towards the origin, the amount can be reduced step by step. For each step, the minimum distance d min The gradual reduction can be carried out in equal or different steps. In particular, the magnitude of the optimal acceleration vector a→optimal be iteratively reduced until the minimum distance d minthe limit of the near collision is reached. In embodiments, the limit of the near collision can be determined by the min > 0 and d min ≤ ε. Here, "ε" can define a near-collision distance ε at which a collision between the vehicle 1 and the collision object 3 is just barely avoided. In embodiments, the near-collision distance ε can be between 1 mm and 1000 mm, preferably between 10 mm and 500 mm, particularly preferably between 50 mm and 100 mm. In particular, the near-collision distance ε can also be <1000 mm, <500 mm, <100 mm, <50 mm, <10 mm, or <1 mm.

[0026] In some embodiments, the criticality can be KRIT optimal the optimal trajectory τ optimal by forming the ratio of the magnitude of the critical acceleration vector a→crit to the magnitude of the optimal acceleration vector a→optimal This means that: KRIToptimal=|a→crit||a→optimal|

[0027] The proposed computer-implemented method is particularly advantageous in subcritical situations, i.e., situations where a collision can be avoided. The value of CRIT can vary from 0 to 1. The lower the value of CRIT, the lower the proportion of the maximum usable vehicle capability that is required to prevent a collision. If the value of a→crit If the criticality value is 0, no evasive maneuver is necessary. The higher the criticality value (CRIT), the greater the minimum percentage of the vehicle's maximum usable capability is required to avoid a collision.

[0028] In case none of the calculated avoidance trajectories τ i to a minimum distance d min> 0, a collision cannot be prevented. This means that in such a driving situation, the min also for the optimal trajectory τ opti-mal negative. In a first variant of the computer-implemented method 100, the criticality KRIT can be limited to 1 in such a case. In a second variant of the computer-implemented method 100, the criticality KRIT can also be greater than 1 in such a case. Analogous to the reduction of the magnitude of the optimal acceleration vector described above, a→optimal, In this second variant, the magnitude of the optimal acceleration vector a→optimal up to a critical acceleration vector a→crit be increased, where a critical trajectory (τ krit ) is determined at which the minimum distance (d min) reaches the threshold of a near-collision again. Thus, even for critical cases where a collision is unavoidable, a risk assessment can be performed based on the maximum usable vehicle capability.

[0029] In embodiments of the computer-implemented method 100 according to FIG. 1, for several, in particular for all, of the calculated plurality of avoidance trajectories τ i a corresponding acceleration vector a→i, a critical acceleration vector a→i,crit, and a corresponding criticality KRIT i Furthermore, based on a comparison of the determined criticalities KRI-T optimal , CRIT i an optimized trajectory τ optimiert The optimized trajectory τ optimiert the trajectory τ ibe determined as having the lowest criticality value. For the determined criticalities KRIT, the following applies analogously to the above explanations: KRITi=|a→i,krit||a→i|

[0030] In embodiments, the computer-implemented method 100 may include an action step for displaying the optimal trajectory τ optimal and / or their criticality KRIT optimal In embodiments, the display may include an acoustic output, a visual output, and / or a vibratory output. In embodiments, the computer-implemented method 100 may include an action step for autonomously or semi-autonomously controlling an evasive maneuver of the at least partially automated vehicle 1 according to the optimal trajectory τ optimal In embodiments, the specific criticality can be KRIT, KRIT i , CRIT optimalused to test and / or validate an at least partially automated driving function system 2 of the at least partially automated vehicle 1. For example, approval can be denied above a certain criticality value, for example, above a value of CRIT = 0.7 or greater.

[0031] In this regard, Fig. 3 schematically shows an at least partially automated vehicle 1 and a referenced collision object 3. The at least partially automated vehicle 1 can be equipped with an at least partially automated driving function system 2. The at least partially automated driving function system 2 can comprise a sensor system and / or receive sensor data from a sensor system. The sensor system can be designed to acquire vehicle data and environmental data. For example, the sensor system can comprise long- and medium-range environmental sensors, for example based on long-range radar (LRR), lidar, mid-range radar (MRR), short-range radar (SRR), ultrasound, video, or the like. Information about the referenced collision object 3 can be obtained via the sensor system. The at least partially automated driving function system 2 can be designed to execute the computer-implemented method 100 described above.The at least partially automated vehicle 1 moves at a current speed. v→ in the longitudinal direction (along the arrow v→). An example is Fig. 3 a calculated avoidance trajectory τ ishown, by which a collision with the collision object 3 can be prevented. It is understood that the collision object 3 or its position can be viewed statically or dynamically. Various approaches for the dynamic observation of a collision object are known to those skilled in the art. For example, a position or movement (i.e., change in position) of the collision object 3 can be projected into the future. The projection of the movement of the collision object 3 can, for example, be carried out with a constant velocity vector. In other examples, the projection of the movement of the collision object 3 can include a change in the velocity vector (e.g., by changing the direction and / or the magnitude of the velocity vector) of the collision object 3, in particular a change towards the trajectory of the vehicle 1. In addition, the trajectory τ i corresponding minimum distance dmin As described above, the computer-implemented method 100 can be carried out during operation of the at least partially automated vehicle 1. For example, the illustrated avoidance trajectory τ i the optimal trajectory τ optimal The computer-implemented method 100 can cause the at least partially automated vehicle 1 to output steering and / or braking signals corresponding to the optimal trajectory τ optimal Alternatively or additionally, a warning and / or a recommendation can be issued via a user interface. Alternatively or additionally, the method 100 can be used to test and / or validate an at least partially automated driving function system 2 of the at least partially automated vehicle 1.

[0032] Furthermore, a computer system is disclosed which is designed to carry out the method 100 for determining a criticality KRIT of the evasive behavior of an at least partially automated vehicle 1 which is traveling at a current speed v→ to carry out while moving. The computer system can comprise at least one processor and / or at least one main memory. The computer system can further comprise a (non-volatile) memory. In examples, all steps of the method 100 can be carried out by the computer system. In some examples, individual steps of the method 100 can be carried out by the computer system. Optionally, results of individual method steps that are not carried out by the computer system can be received by the computer system. In embodiments, the computer system can be included in the at least partially automated driving function system 2 described above.

[0033] Furthermore, a computer program is disclosed which is designed to implement the method 100 for determining a criticality KRIT of the evasive behavior of an at least partially automated vehicle 1 which is traveling at a current speed v→ The computer program can be in interpretable or compiled form, for example. It can be loaded (even in parts) into the RAM of a computer for execution, e.g., as a bit or byte sequence. In embodiments, the computer program can be stored in a memory of the at least partially automated driving function system 2 described above.

[0034] Further disclosed is a computer-readable medium or signal that stores and / or contains the computer program or at least a portion thereof. The medium may, for example, comprise one of RAM, ROM, EPROM, HDD, SDD, etc., on / in which the signal is stored.

Claims

[1] Computer-implemented method (100) for determining a criticality (KRIT) of the evasive behavior of an at least partially automated vehicle (1), wherein the at least partially automated vehicle (1) is traveling at a current speed (v→) moved, the method (100) comprising the following steps: - Calculating (110) a plurality of avoidance trajectories (τ i ) of the vehicle (1) with respect to a referenced collision object (3) by combining different longitudinal accelerations (12) with different lateral accelerations (14); - Determine (120) an optimal trajectory (τ optimal ) from the majority of avoidance trajectories (τ i ), where a minimum distance (d min ) between the vehicle and the collision object (3) is maximum; - Determining (130) a critical acceleration vector (a→crit) a corresponding critical trajectory (τ krit ), where the minimum distance (d min ) reaches a near-collision threshold; - Determine (140) a criticality (CRIT optimal ) of the optimal trajectory (τ optimal ) based on the critical acceleration vector (a→crit) and one to the optimal trajectory (τ optimal ) corresponding optimal acceleration vector (a→optimal). [2] The computer-implemented method (100) of claim 1, wherein determining the critical acceleration vector (a→crit) reducing the magnitude of the optimal acceleration vector (a→optimal) up to the critical acceleration vector (a→crit) where the minimum distance (d min ) reaches the limit of a near collision. [3] Computer-implemented method (100) according to any one of the preceding claims, wherein the criticality (CRIT) of the optimal trajectory (τ optimal ) by forming the ratio of the magnitude of the critical acceleration vector (a→crit) to the magnitude of the optimal acceleration vector (a→optimal) is determined. [4] A computer-implemented method (100) according to any one of the preceding claims, wherein the near-collision boundary region is defined by (d min > 0) and (d min ≤ ε), where (ε) defines a near collision distance at which a collision is just avoided. [5] Computer-implemented method (100) according to any one of the preceding claims, wherein combining different longitudinal accelerations (12) with different lateral accelerations (14) comprises combinations of maximum usable braking capability and maximum usable steering capability according to a predetermined maximum friction force model, in particular a predetermined Kamm circle model (10). [6] A computer-implemented method (100) according to any one of the preceding claims, wherein the minimum distance (d min ) between the vehicle (1) and the collision object (3) for each avoidance trajectory (τ i ) is determined, and optionally, where the respective minimum distances (d min ) of the majority of avoidance trajectories (τ i ) are compared with each other, and the avoidance trajectory (τ optimal ) of the majority of avoidance trajectories (τ i ) with the largest minimum distance (dmin ) as optimal avoidance trajectory (τ optimal ) is selected. [7] Computer-implemented method (100) according to any one of the preceding claims, - where for several, in particular for all, of the calculated plurality of avoidance trajectories (τ i ) a corresponding acceleration vector (a→i), a critical acceleration vector (a→i, crit), and a corresponding criticality (CRIT i ) and optionally - based on a comparison of the determined criticalities (CRIT optimal , CRIT i ) an optimized trajectory (τ optimiert ) is determined. [8] A computer-implemented method (100) according to any one of the preceding claims, further comprising an action step: - to display the optimal trajectory (τ optimal ) and / or their criticality (CRIT optimal ), and / or - for the autonomous or semi-autonomous control of an evasive maneuver of the at least partially automated vehicle (1) according to the optimal trajectory (τ optimal ). [9] A computer-implemented method (100) according to any one of the preceding claims, wherein the determined criticality (CRIT i , CRIT optimal ) is used to test and / or secure an at least partially automated driving function system (2) of the at least partially automated vehicle (1). [10] Computer system designed to carry out the method (100) for determining a criticality (KRIT) of the evasive behavior of an at least partially automated vehicle (1) according to any one of the preceding claims. [11] Computer program comprising instructions which, when the computer program is executed by a computer system, cause the computer system to carry out the method (100) for determining a criticality (KRIT) of the evasive behavior of an at least partially automated vehicle (1) according to any one of claims 1 to 9. [12] A computer-readable medium or signal storing and / or containing the computer program of claim 11. [13] Method (300) for setting up an at least partially automated driving function system (2), wherein the criticality (KRIT) determined according to any one of the preceding claims 1 to 9 of the computer-implemented method (100) is taken into account when setting up at least partially automated driving functions, in particular when setting up rules for evasive maneuvers.

Images