METHOD FOR DETERMINING A TRAJECTORY OF AN AT LEAST PARTIALLY ASSISTED OPERATED MOTOR VEHICLE, COMPUTER PROGRAM AND ASSISTANCE SYSTEM

Abstract

Technologies and techniques for determining a trajectory of an assisted-operated motor vehicle. At least one object is detected in an environment of the motor vehicle and at least one uncertainty with respect to the object is determined. A future environment with the object is predicted via an electronic computing device, as a function of the detected environment and the detected object, wherein a risk value for a planned trajectory is determined on the basis of a collision probability. A most probable impact constellation and accident severity for the most probable impact constellation is determined, wherein the collision probability and the accident severity is weighted in a risk value, and wherein the trajectory is determined as a function of the determined risk value.

Claims

1. A method for determining a trajectory of an at least partially assisted-operated motor vehicle via an assistance system of the motor vehicle, comprising: detecting, via an environment detection device, at least one object in an environment of the motor vehicle; determining, via an electronic computing device, at least one uncertainty with respect to the detected object; predicting, via the electronic computing device, a future environment of the object as a function of the detected environment and the detected object; and determining, via the electronic computing device, a risk value for a planned trajectory on the basis of a collision probability, a determined most probable impact constellation, and a determined accident severity for the most probable impact constellation, wherein the collision probability and the accident severity are weighted in the risk value, and wherein the planned trajectory is determined as a function of the determined risk value.

2. The method according to claim 1, further comprising determining an evasive maneuver or a mitigation maneuver as the trajectory with respect to the object.

3. The method according to claim 2, wherein the planned trajectory comprises an at least partially assisted intervention in a longitudinal acceleration device and/or in a transverse acceleration device of the motor vehicle.

4. The method according to claim 1, wherein the at least one uncertainty comprises one or more of: a position uncertainty of the object, a position uncertainty of the motor vehicle, a pose uncertainty of the object, a pose uncertainty of the motor vehicle, a parameter uncertainty of the object, and/or a parameter uncertainty of the motor vehicle.

5. The method according to claim 1, wherein determining the risk value comprises determining the risk value on an instantaneous level for the future environment.

6. The method according to claim 1, wherein determining the risk value comprises determining the risk value in real time.

7. The method according to claim 1, wherein determining the most probable impact constellation comprises processing a Minkowski difference and a linear transformation.

8. The method according to claim 1, further comprising processing a distribution of the accident severity during the determining of the risk value.

9. An assistance system for an at least partially assisted-operated motor vehicle for determining a trajectory, comprising: an environment detection device, for detecting at least one object in an environment of the motor vehicle; an electronic computing device, for determining at least one uncertainty with respect to the detected object; wherein the electronic computing device is configured to predict a future environment of the object as a function of the detected environment and the detected object, and wherein the electronic computing device is configured to determine a risk value for a planned trajectory on the basis of a collision probability, a determined most probable impact constellation, and a determined accident severity for the most probable impact constellation, wherein the collision probability and the accident severity are weighted in the risk value, and wherein the planned trajectory is determined as a function of the determined risk value.

10. The assistance system according to claim 9, further comprising determining an evasive maneuver or a mitigation maneuver as the trajectory with respect to the object.

11. The assistance system according to claim 10, wherein the planned trajectory comprises an at least partially assisted intervention in a longitudinal acceleration device and/or in a transverse acceleration device of the motor vehicle.

12. The assistance system according to claim 9, wherein the at least one uncertainty comprises one or more of: a position uncertainty of the object, a position uncertainty of the motor vehicle, a pose uncertainty of the object, a pose uncertainty of the motor vehicle, a parameter uncertainty of the object, and/or a parameter uncertainty of the motor vehicle.

13. The assistance system according to claim 9, wherein determining the risk value comprises determining the risk value on an instantaneous level for the future environment.

14. The assistance system according to claim 9, wherein determining the risk value comprises determining the risk value in real time.

15. The assistance system according to claim 9, wherein determining the most probable impact constellation comprises processing a Minkowski difference and a linear transformation.

16. The assistance system according to claim 9, further comprising processing a distribution of the accident severity during the determining of the risk value.

17. A computer program product comprising non-transitory program code, which, when executed, causes an assistance system for a motor vehicle to: detect at least one object in an environment of the motor vehicle; determine at least one uncertainty with respect to the detected object; predict a future environment of the object as a function of the detected environment and the detected object; and determine a risk value for a planned trajectory on the basis of a collision probability, a determined most probable impact constellation, and a determined accident severity for the most probable impact constellation, wherein the collision probability and the accident severity are weighted in the risk value, and wherein the planned trajectory is determined as a function of the determined risk value.

18. The computer program product according to claim 17, further comprising determining an evasive maneuver or a mitigation maneuver as the trajectory with respect to the object, wherein the planned trajectory comprises an at least partially assisted intervention in a longitudinal acceleration device and/or in a transverse acceleration device of the motor vehicle.

19. The computer program product according to claim 17, wherein the at least one uncertainty comprises one or more of: a position uncertainty of the object, a position uncertainty of the motor vehicle, a pose uncertainty of the object, a pose uncertainty of the motor vehicle, a parameter uncertainty of the object, and/or a parameter uncertainty of the motor vehicle.

20. The computer program product according to claim 17, wherein determining the risk value comprises one of determining the risk value on an instantaneous level for the future environment, or determining the risk value in real time.

Description

BRIEF DESCRIPTION OF THE DRAWINGS

[0030] In the following, various aspects of the present disclosure are described. In the figures:

[0031] FIG. 1 a schematic top plan view of a motor vehicle including an assistance system according to some aspects of the present disclosure;

[0032] FIG. 2 a schematic flow diagram according to some aspects of the present disclosure;

[0033] FIG. 3 a schematic top plan view of a first traffic situation according to some aspects of the present disclosure; and

[0034] FIG. 4 a schematic top plan view of a further traffic situation according to some aspects of the present disclosure.

DETAILED DESCRIPTION

[0035] The embodiments described below are preferred embodiments of the present disclosure. In the embodiments, the components described each represent individual features of the present disclosure which are to be considered independently of one another and which each also further the present disclosure independently of one another and thus are also to be regarded as a component of the present disclosure individually or in a combination other than that shown. Furthermore, the described embodiments can also be expanded by further ones of the already described features of the present disclosure.

[0036] In the figures, functionally identical elements are each provided with the same reference signs.

[0037] FIG. 1 shows a schematic top plan view of a motor vehicle 1 including an assistance system 2 according to some aspects of the present disclosure. In this example, the motor vehicle 1 is at least partially assisted-operated. In particular, the assistance system 2 can thus intervene, for example, in a lateral control or in a longitudinal control of the motor vehicle 1. The motor vehicle 1 may also be fully automated. The assistance system 2 has at least one environment detection device 3 and an electronic computing device 4. An environment 5 and a further object 6, in this case a further motor vehicle, can be detected by means of the environment detection device 3. In particular, the object pose of the object 6 can be detected. While detecting the further motor vehicle, in particular, an uncertainty 7 may occur. This uncertainty 7 is, in particular, a pose uncertainty of the further motor vehicle or a position uncertainty of the further motor vehicle, for example. The motor vehicle 1 may also have a position uncertainty and a pose uncertainty. Furthermore, a parameter uncertainty, for example, regarding the dimension of the contour of the object 6 or of the motor vehicle 1, can also be taken into account.

[0038] FIG. 2 shows a schematic flow diagram according to according to some aspects of the present disclosure. In this example, it is shown that by means of the assistance system 2 a trajectory 8 (FIG. 3) of the at least partially assisted motor vehicle 1 can be determined. For this purpose, in a first step S1 at least the object 6 in the environment 5 of the motor vehicle 1 is detected by means of the environment detection device 3 and the uncertainty 7 with respect to the object 6 is determined by means of the electronic computing device 4 of the assistance system 2. A future environment 5 with the object 6 is predicted as a function of the detected environment 5 and the detected object 6 in a second step S2 by means of the electronic computing device 4. In a third step S3, a risk value R (FIG. 3) for a planned trajectory 8a, 8b, 8c (FIG. 4) is determined on the basis of a collision probability 9 determined by means of the electronic computing device 4 and a determined most probable impact constellation 10 and a determined accident severity 11 for the most probable impact constellation 10, wherein the collision probability 9 and the accident severity 11 are weighted in the risk value R on the basis of a weight 12, and wherein the trajectory 8 is determined as a function of the determined risk value R. The determination of the trajectory 8 is shown in the present case in particular by the fourth step S4. In a fifth step S5, in turn, the determined emergency maneuver, i.e. the trajectory 8, can then be performed. Then, in a sixth step S6, a readjustment can be carried out, in which it is possible to go back to the first step S1, whereby a continuous optimization of the vehicle action can be realized.

[0039] It may, in particular, be provided that an avoidance maneuver 8a (FIG. 4) with respect to the object 6 or a mitigation maneuver 8c (FIG. 4) with respect to the object 6 is determined as the trajectory 8. Furthermore, regarding the trajectory 8, an at least partially assisted intervention in a longitudinal acceleration device 14 and/or a lateral acceleration device 15 of the motor vehicle 1 is possible.

[0040] As mentioned above, a position uncertainty of the object 6 and/or a position uncertainty of the motor vehicle 1 can be determined as the uncertainty 7. Furthermore, a pose uncertainty of the object 6 and a pose uncertainty of the motor vehicle 1 can also be determined. Furthermore, the parameter uncertainty can also be determined.

[0041] It is provided herein that determining the trajectory 8, in particular, determining the risk value R, is carried out in real time.

[0042] In the example of FIG. 3 a first traffic situation of the motor vehicle 1 is described. The motor vehicle 1 drives straight ahead and “suddenly” the object 6 appears as a further motor vehicle, for example, due to a parking maneuver of the further motor vehicle or due to the further motor vehicle taking the right of way, as well as a further object 16 as once again a further motor vehicle and as oncoming traffic. The one further motor vehicle remains in the same lane as the motor vehicle 1 and thus impedes continued driving. In this case, the last point for braking with collision avoidance has already been exceeded for motor vehicle 1. Nevertheless, the collision with the object 6 could be avoided by taking evasive action. In this embodiment example, oncoming traffic hinders an unobscured or safe evasive maneuver 8a. For an appropriate simulation of such a traffic situation, a test drive can be repeated with changing but predefined uncertainty level with respect to the occurring objects 6, 16.

[0043] The scenario requires selecting between two unfavorable options. On the one hand, braking would mitigate the collision with object 6 but also cause a certain accident, thus forming a mitigation maneuver 8c. On the other hand, the evasive maneuver offers the chance of collision avoidance with both objects 6, 16, but if it fails, the collision occurs with even higher severity because of the high relative speed with respect to the oncoming traffic. In other words, if it is possible to predict a safe avoidance maneuver 8a, it is preferred to collision mitigation. Conversely, if collision avoidance is not possible, mitigation is preferred. The decision depends on the potential accident severity 11 and its uncertainty 7 the risk value R. In other words, a low uncertainty 7 allows an informed decision for the avoidance maneuver 8a. However, if the uncertainty 7 is too high, collision mitigation is chosen.

[0044] In the following, the existence and classification uncertainties are considered the most probable. In other words, an object 6 either exists or does not exist, and in the case of existence, only one object class is assigned. Therefore, an equation to determine the risk value simplifies to:

[00001]$\begin{array}{cc}R\left(\pi \right)=\underset{:=R_{f,t_i}}{\underbrace{\underset{\begin{array}{cc}& \end{array}}{{\Sigma }_{i=1}^{N_{\mathrm{inst}}}{\Sigma }_{k=1}^{N_H}{\Sigma }_{j=1}^{N_H}p\left(h_k^j\right).Math.R\left({\mathrm{EK}}_{i,k}^j\right)}}}& \left(1\right)\end{array}$

Moreover, the uncertainties of length l.sub.E,k and width w.sub.E,k are neglected. With respect to object 6, this is based on the assumption that the geometric impact is superimposed on the object pose. The ego vehicle length l.sub.E and width w.sub.E are assumed to be known based on design information independent of the planning process.

[0045] The accident severity 11 of a partial state is modeled as a random variable with a sample space specifying all possible object constellations between the ego-vehicle and a target object (accident-free, accident constellation, etc.) with the probability and probability density function. Accordingly, the expected value R can be calculated with:

[00002]$\begin{array}{cc}R\left(\mathrm{EK}\right)=E\left(\Psi \right)={\int }_{{\Omega }^{\mathrm{voc}}}\Psi \left(\omega \right).Math.\mathrm{dP}\left(\omega \right)\stackrel{\mathrm{discretisation}}{}E\left(\Psi \right)={.Math.}_n{\psi }_n.Math.p_n={\int }_{{\mathrm{\omega ϵ\Omega }}^{\mathrm{voc}}}\psi \left(\omega \right){𝓅}_{{\Omega }^{\mathrm{VOC}}}\left(\omega \right)d\omega & \left(2\right)\end{array}$

[0046] In this way, the severity level of incremental damage Y(w) and their incremental probabilities of occurrence dP(w) are combined, namely the incremental risk dR, which in turn is aggregated to the risk R(EK) of the partial state. The severity level Y is determined according to the predictive models h with the input vector according to the environmental model. Accordingly, the risk R(EK) is determined with the joint probability density function, while the geometric state of the motor vehicle 1 and the object 6, and a parameter vector can be modeled in a mutually stochastically independent manner

[00003]$$\begin{gathered} \begin{array}{cc}R\left(\mathrm{EK}\right)=E\left(\Psi \right)={\int }_{{\tilde{g}}_E}{\int }_{{\tilde{g}}_K}{\int }_{p_E}{\int }_{p_K}\eta \left(z\right){𝓅}_{X_E,Y_E,{\Phi }_E}\left(x_E,y_E,{\phi }_E\right){𝓅}_{X_K,Y_K,{\Phi }_K}\left(x_K,y_K,{\phi }_K\right).Math. \\ {𝓅}_{p_E}\left(p_E\right){𝓅}_{p_K}\left(p_K\right)d{\tilde{g}}_{E}d{\tilde{g}}_K{\mathrm{dp}}_E{\mathrm{dp}}_K& \left(3\right)\end{array} \end{gathered}$$

[0047] Assuming cooperative behavior between average traffic participants, the independence between motor vehicle 1 and object 6 overestimates the risk R(EK) and is therefore a conservative approximation. However, due to the normal distributions of Z, there is no analytical solution. In addition, random calculations or numerical integrations are not suitable due to the computation time requirements. Therefore, the following section deals with suitable approximations to the equation shown above.

[0048] A severity level Y>0 occurs only when an accident occurs. Therefore, the equation is separated into the geometric variables related to the probability of an accident (I) and damage (II):

[00004]$\begin{array}{cc}2R\left(\mathrm{EK}\right)=E\left(\Psi \right)={\int }_{{\tilde{g}}_E}{\int }_{{\tilde{g}}_K}{\eta }_{\mathrm{ind}}\left(z\right){𝓅}_{X_E,Y_E,{\Phi }_E}\left(x_E,y_E,{\phi }_E\right){𝓅}_{X_K,Y_K,{\Phi }_K}\left(x_K,y_K,{\phi }_K\right).Math.[{\int }_{{\tilde{g}}_E.Math.\mathrm{coll}}{\int }_{{\tilde{g}}_K.Math.\mathrm{coll}}{\int }_{p_E}{\int }_{p_K}\eta \left(z\right).Math.& \left(I\right)\left(4\right)\end{array}$$\begin{array}{cc}{𝓅}_{{\tilde{G}}_E.Math.\mathrm{coll}}\left({\tilde{g}}_E.Math.\mathrm{coll}\right){𝓅}_{{\tilde{G}}_K.Math.\mathrm{coll}}\left({\tilde{g}}_K.Math.\mathrm{coll}\right).Math.{𝓅}_{P_E}\left(p_E\right){𝓅}_{P_K}\left(p_K\right)d{\tilde{g}}_{E}d{\tilde{g}}_K{\mathrm{dp}}_E{\mathrm{dp}}_K]d{\tilde{g}}_{E}d{\tilde{g}}_K& \left(\mathrm{II}\right)\end{array}$$\begin{array}{cc}{\eta }_{\mathrm{ind}}\left(z\right)=\{\begin{array}{cc}1& \vartheta \left(x_E,y_E,{\phi }_E,l_E,w_E,\right).Math.\vartheta \left(x_K,y_K,{\phi }_K,l_K,w_K\right)\ne ⌀\\ 0& \vartheta \left(x_E,y_E,{\phi }_E,l_E,w_E,\right).Math.\vartheta \left(x_K,y_K,{\phi }_K,l_K,w_K\right)=⌀\end{array}& \left(5\right)\end{array}$

[0049] This function reports a collision when the vehicle contour regions J overlap, depending on the geometric states.

[0050] To distinguish different collision configurations, the second part II) of the above equation indicates the accident severity 11 for each accident constellation. Again, due to modeling with normal distributions, there is no closed-form solution. Moreover, the severity prediction model h(Z) is nonlinear and deforms the shape of the normal distribution. On the one hand, the integral II) could be solved by permutation over the collision configurations with y>0. As a result, the equation changes to:

[00005]$\begin{array}{cc}R\left(\mathrm{EK}\right)=E\left(\Psi \right)=\underset{I)}{\underbrace{P\left(C\right)}}.Math.\underset{\mathrm{II})}{\underbrace{{.Math.}_{n=1}^{N_{\mathrm{coll}.\mathrm{const}.}}P\left(z_n|\mathrm{coll}\right).Math.\eta \left(z_n\right)}}& \left(6\right)\end{array}$

[0051] The accuracy here depends directly on the discretization. Theoretically, with an infinitely small step size and infinitely many collision constellations, respectively, the expected values of the severity E(Y) and thus the risk R(EK) can be calculated exactly. However, even smaller numbers require a high computation time, which is not suitable for real-time applications. Alternatively, uncertainty propagation through the severity level model h provides a distribution of the severity level independent of arbitrary discretizations. Moreover, it allows for the evaluation of different quantiles yp-quantile of the severity level in addition to the expected value.

[0052] This approximation could be used for additional margins of safety. Since sample-based approaches are not suitable for real-time applications, the shape of a normal distribution requires linear uncertainty propagation to avoid deforming the distribution:

[00006]$\begin{array}{cc}\begin{array}{ccc}{\Psi }_{\mathrm{coll}}=\eta (\mathrm{{\rm Z}│}|\mathrm{coll}& \stackrel{\mathrm{linearisation}}{}& {\tilde{\Psi }}_{\mathrm{coll}}\sim 𝒩\left({\mu }_{\tilde{\Psi }},{\Sigma }_{\tilde{\Psi }}\right)\mathrm{with}\\ {\mu }_{\tilde{\Psi }}=\eta \left({\mu }_{{\rm Z}|\mathrm{coll}}\right)& & \end{array}& \left(7\right)\end{array}$

Here, ∇(h) indicates the Jacobi matrix of the severity level prediction model evaluated at the point mZ.

[0053] The main objective is to estimate the risk R and the expected value, respectively. Thus, the equation (4) changes to:

[00007]$\begin{array}{cc}R\left(\mathrm{EK}\right)=E\left(\Psi \right)=\underset{I)}{\underbrace{P\left(C\right)}}.Math.\underset{\mathrm{II})}{\underbrace{E\left({\Psi }_{\mathrm{coll}}\right)}}& \left(8\right)\end{array}$

Assuming symmetric, flattening input distributions, such as with normal distribution, the expected value is equal to the severity level of the most probable collision configuration. This leads to the risk estimate:

[00008]$\begin{array}{cc}R\left(\mathrm{EK}\right)=E\left(\Psi \right)=\underset{I)}{\underbrace{P\left(C\right)}}.Math.\underset{\mathrm{II})}{\underbrace{{\psi }_{\mathrm{Pmax}}}}& \left(9\right)\end{array}$

[0054] In summary, the risk assessment stated above is used for each instantaneous partial state zEK present in the assessment. However, this serves only to simplify the concept according to the present disclosure and is by no means to be regarded as conclusive. Other risk assessments presented may also be used.

[0055] The separation into collision probability 9 and accident severity 11 provides for the separation between preventive driving and emergency maneuvers. For regular driving, accident severity 11 is less important compared to collision probability 9. Here, collision probability 9 ensures safe driving. In contrast, severity distinguishes different collision configurations in emergency driving maneuvers.

[0056] Furthermore, a short distance to the possible collision results in a high resolution of a Markov's decision process regarding the collision configurations. The most probable accident configuration determines the accident severity 11. Here an assumption is made about the shape of the distribution. The input shape must be symmetrically flattening to be valid for exact calculations. Otherwise, it is a slight approximation compared to the inaccuracies of the severity prediction model, which allows real-time applications due to the smaller number of calculations. The normal distributions of the inputs represent a special case of a symmetrically flattening distribution. In addition, the most probable geometric collision configuration is derived using mechanisms similar to those used in the calculation of the collision probability 9, and thus with low computational effort. All other input parameters for determining the accident severity are not conditional and can thus be derived directly from the respective distributions.

[0057] The separation into collision probability 9 according to I) and accident severity 11 according to II) also allows for the use of established nonlinear methods for efficient estimation of collision probability 9. Furthermore, the severity distribution could be derived by Gaussian uncertainty propagation. As a result, detailed safety requirements can be set and are thus open for future standards.

[0058] According to the bivariate collision probability 9 method, the most probable geometric collision configuration can be directly identified. The equipotentials of the standardized normal distribution are concentric circles around the point of origin. Therefore, the shortest distance to the collision area, which is described herein terms of a Minkowski difference, indicates the most probable geometric collision configuration. It is either a vertex directly or a point between the two shortest vertices. The Minkwoski difference can be further subdivided into subranges representing collision configurations that permute the geometric features of the gravity prediction model h. Finally, the mapping correlates with the untransformed collision domain due to the linear relationship and thus the input for gravity prediction is obtained.

[0059] The example of FIG. 4 shows a further traffic scenario according to some aspects of the present disclosure. This scenario is used to describe the mitigation performance of the risk-based planner. FIG. 4 describes that the motor vehicle 1 is driving straight ahead, when “suddenly” a potential collision object, in this case the object 6, appears. This can, for example, occur when object 6 exits a parking space. Consequently, motor vehicle 1 must perform an emergency maneuver. Depending on the object distance, different options are available. The underlying speeds cause the last point to brake to be passed before the last point to steer for accident avoidance. If the distance is great enough, the motor vehicle 1 is free to choose an appropriate evasive maneuver 8a to avoid a collision based on the risk assessment. If the distance decreases, a collision can only be avoided by steering. If the object 6 appears very suddenly, avoidance is no longer possible, but a mitigation maneuver 8c can still reduce impact severity. Furthermore, motor vehicle 1, which has already “moved on”, is provided with the reference sign 13 in FIG. 4. Reference sign 8b indicates in particular the collision trajectory or the collision maneuver without corresponding intervention by the assistance system.

[0060] Furthermore, FIG. 4 shows a modification of the scenario presented above. Here, two objects 6, 16 appear in front of the motor vehicle 1. The further object 16 impedes a collision-free emergency evasive maneuver. Meaning, a collision becomes unavoidable earlier than in the scenario variant before. But even in this case, the motor vehicle 1 still has the option to reduce the collision severity to ensure maximum safety.

[0061] While the further object 6 can legally drive on the adjacent lane, the object 6 disregards the right of way and drives out behind the obscuring parked cars, for example. In addition, these parked cars make a collision maneuver to the right difficult.

[0062] Various metrics can be defined to quantify the degree of accident severity 11 of crash consequences. The parameters vary depending on the field of research and the specific area of interest. In general, an accident exists when unintended forces are applied to the vehicle body, resulting in adverse health effects or damage. In addition, a variety of influences act on the crash outcome, making it difficult to objectively quantify the accident severity 11, especially with few individual values. For example, the same technical accident sequence can lead to completely different short-term effects for vital and non-vital traffic participants, with even more uncertain long-term effects. Therefore, the range of accident severity is divided into four groups below,

[0063] A first group constitutes vehicle and occupant loading, a second group the technical severity, a third group the injury severity, and a fourth group the long-term effects.

[0064] Based on advantages and disadvantages, the particular application and requirements determine the appropriate severity level metric. In addition, the availability of data leads to necessary approximations.

[0065] The technical accident severity 11 quantifies the mechanical vehicle loading due to force that result in acceleration a(t), velocity v(t) and deformation s(t) over time.

[0066] The technical accident severity 11 depends, among other things, on the type of collision objects (masses, shapes, compatibility, etc.), the speed and the impact position. Characteristic values of kinematics are used to indicate the severity by means of single values. The key here is, where and how the data is obtained. While FEM simulations, crash tests, and event recorders provide detailed information about the crash history (e.g., a(t), v(t), and s(t)), police and accident investigators record the incident retrospectively by finding the vehicle only in the rest position. Nevertheless, various accident severity level metrics have been established over the years. For example, the deformation energy ΔT is obtained by reconstructing the force by way of the intrusion. Since the intrusion is measured after the impact, it only indicates the plastic energy exchange. The Energy Equivalent Speed (EES) relates the deformation energy ΔT as kinetic energy to the vehicle mass m: ΔT=0.5mEES.sup.2. The reconstruction of the accident includes the determination of the velocities at impact, denoted vrel, as well as the velocity change during the crash Av.

[0067] With respect to the vehicle itself, the external force loads the vehicle body and thus indirectly the occupants. Consequently, the crash impulse a(t) and the intrusion s(t) are the main causes of injuries. To mitigate the damage, energy is extracted from the passenger compartment by means of vehicle deformation.

[0068] In addition, the restraint systems are designed to distribute the load on the occupants over the crash time according to the human load limits. Nevertheless, strong impacts can still act on the individual occupants. These are measured, for example, by the Head Injury Criterion (HIC) or Neck Injury Criterion (NIC), which indicate the acceleration of the respective body region over a certain time interval. The same is true for other traffic participants, such as Vulnerable Road Users (VRU). The only difference is probably the lack of appropriate impact protection.

[0069] Injury severity depends on occupant-related characteristics, such as vitality, height, or gender, and occupant position in the vehicle, in addition to force and restraint system. The Abbreviated Injury Scale (AIS) is a commonly used metric in accident research to indicate and compare the medical severity level.

[0070] The AIS assesses the lethality of individual injuries. The Maximum AIS (MAIS) represents these individual injuries of body regions or the entire person by their maximum value. Alternatively, the Injury Severity Score (ISS) aggregates the most severe traumas of three body regions quadratically.

[0071] In addition, long-term effects can be expressed in monetary values, such as vehicle damage and medical costs, or in human-related characteristics, such as convalescence, survival probability, or lethality rate.

[0072] In summary, based on this overview, there is no all-in-one solution to express the damage of a collision. Rather, the application must determine the appropriate severity level metric. Due to automated driving with human traffic participants, the goal in the aforementioned embodiments is to protect traffic participants based on ethical guidelines. Therefore, injury severity has been considered as a metric. However, injury severity level is highly individual, which makes objective crash assessment difficult, and is very difficult, if not impossible, to predict with sufficient accuracy in real time. Furthermore, the selected application example in this embodiment is based on the impact of the vehicle structure. It does not necessarily require injury severity and can be expressed by vehicle crash dynamics. Therefore, in these embodiments, the technical accident severity 11 is chosen to represent the accident damage. Advantageous for the presented method is selecting the technical accident severity 11 through Δv.

[0073] Even in the case of a head-on accident, for example, several parameters can be used to indicate severity. Furthermore, the individual vehicle dynamics a(t), v(t) and s(t) are convertible or redundant, respectively, so that it seems reasonable to reduce considerations. The restraint systems are mainly dependent on deceleration and speed. Deformation does not give any information about stopping behavior and possible multiple collisions.

[0074] Furthermore, the acceleration signals will be noisy in most cases when measured, for example, by an event recorder or in FEM simulations. Additionally, it must be mentioned that a single value is a rough approximation and in this case neglects important temporal features, such as maximum or average deceleration

[0075] On the other hand, the value Δv has a long tradition in depth surveys and it has a strong correlation with injury probabilities. The probability of a given MAIS level is related to the recorded Δv value by logistic regression.

NOMENCLATURE LIST AND SYMBOL LIST

[0076] The nomenclature and symbol list below is written in English, since it is the primarily used language in the field of autonomous driving and to stay consisted with the corresponding abbreviations used. It is added for the sake of completeness and serves, in particular, for interpreting and understanding the formulas used in the description. Any abbreviations that are not written out in full and/or formula symbols that are not explained can therefore be taken from the list below. From the following abbreviations and symbol usages, the person skilled in the art can obtain the corresponding notes on calculating the individual formulas or on the corresponding interrelationships. [0077] GDP Gross domestic product [0078] SAE Society of Automotive Engineers [0079] EgoOwn (ego) vehicle [0080] TTC Time-To-Collision [0081] NCAP New Car Assessment Programme [0082] AIS Abbreviated Injury Scale [0083] MAIS Maximal Abbreviated Injury Scale [0084] GIDAS German In-Depth Accident Study [0085] GPU Graphics Processing Unit [0086] FRG Federal Republic of Germany [0087] ADAS Advanced Driver Assistance Systems [0088] IIHS Institute for Highway Safety [0089] ABS Anti-lock Braking System [0090] ESC Electronic Stability Control [0091] AACN Advanced Automatic Crash Notification [0092] eCall Emergency call E911 [0093] TPS Third Party Services [0094] MKB Multi collision brake [0095] ATMS Advanced Traffic Management Systems [0096] DMS Dynamic Message Sign [0097] HMI Human Machine Interface [0098] FE(M) Finit Element (Method) [0099] NASS-CDS National Automotive Sampling System-Crashworthiness [0100] Data System [0101] ICS Inevitable collision state [0102] TM Tunnel Middle [0103] TTX Time to x [0104] TTR Time to react [0105] TTB Time to brake [0106] TTS Time to steer [0107] CA Collision avoidance [0108] CM Collision mitigation [0109] ASIL Automotive Safety Integrity Level [0110] TTCCP Time-to-critical-collision-probability [0111] OEM Original Equipment Manufacturer [0112] MPC Model Predictive Control [0113] RK3 Runge-Kutta integrator third order [0114] AEB Automatic Emergency Braking [0115] VRU Vulnerable Road User [0116] MDP Markov Decision Process [0117] PDF Probability density function [0118] DOF Degree of freedom [0119] CDF Cumulative density function [0120] PCA Principal Component Analysis [0121] COG Center of Gravity [0122] SUV Sports utility vehicle [0123] MSE Mean squared error [0124] TP/FP True/false positives [0125] TN/FN True/false negatives [0126] FFNN Feed forward neuronal network [0127] RF Random forest [0128] CIM Centric impact model [0129] EIM Eccentric impact model Kelvin model. Two masses are connect by a parallel spring and damper [0130] NOC Number of Conflicts [0131] CV Constant velocity [0132] CTR Constant turn rate [0133] LHS Latin hypercube sampling [0134] FS Functional scenario [0135] FES Functional evaluation scenario [0136] LS Logical scenario [0137] CS Concrete scenario [0138] GA Genetic algorithm [0139] GPSA Generalized Pattern Search Algorithm [0140] UTYPGIDAS accident type [0141] HIL Hardware-in-the-loop [0142] INS Inertial Navigation System [0143] GNSS Global Navigation Satellite System [0144] PCM Pre-Crash-Matrix [0145] V2I Vehicle-to-infrastructure [0146] LPTB Last point to brake [0147] LPTS Last point to steer [0148] SIL Software in the loop

[0149] (⋅) Placeholder for a variable [0150] n arbitrary counter [0151] N Absolute number of a finite set [0152] a.sub.b,c Notation means: variable a with the properties b AND c (e.g. ego velocity in longitudinal direction: v.sub.E,long) [0153] a.sub.bc Notation means: variable a with the property b AND variable a with the property c (e.g., velocity v for the ego vehicle E and velocity for the target vehicle K: v.sub.E/K) [0154] alb Event a under the condition b (e.g., accident severity under the condition of collision: I coil) [0155] A bold, capital letter indicates a matrix, or vector of random variables [0156] A,a bold symbol indicates a vector or matrix [0157] P(e) Probability of event [0158] e(⋅)(t) Time variant value (e.g., a(t), v(t), s(t)) [0159] (⋅).sub.long Value in longitudinal direction (e.g., a.sub.long) [0160] (⋅).sub.lat Value in lateral direction (e.g., a.sub.iat) [0161] Pz.sup.(z) Probability density function to the random variable [0162] Z sample [0163] (⋅).sub.E, (⋅).sub.K, (⋅).sub.EK The bidirectional relation between ego vehicle E and one target object K is emphasized by the identifiers E/K/EK [0164] Z˜(μ, σ) The random variable Z is normal distributed with the expected value μ and standard derivation [0165] Z˜(μ, Σ) The random vector Z is normal distributed with the expected value vector μ and covariance matrix Σ [0166] z: Ω.fwdarw. The random variable z maps the sample space Ω to a scalar value of a real number Real numbers [0167] E(⋅) Expected value [0168] (⋅)* Optimal value [0169] (⋅).sub.n, (⋅).sub.t Normal and tangential direction [0170] (⋅).sub.f, (⋅).sub.r Front and rear axle of the non-linear single track model [0171] (⋅) and (⋅)′ Before and afterwards [0172] Δ(⋅) Relative values (e.g., relative pose such as Δx, Δy, and [0173] {circumflex over (()}⋅), (⋅) Estimated value {circumflex over (()}⋅)in relation to the reference value (⋅) [0174] f(⋅) Function in general; [0175] t time [0176] t.sub.0 Time at the moment 0 (begin of a sequence); [0177] t.sub.p, t.sub.i, t.sub.m Different time t discretization levels (t.sub.p: between states s, t.sub.i: reward generation, t.sub.m: integration steps dynamic model) [0178] i Index of time step t.sub.i for the reward generation [0179] t.sub.i(s, a, s′) of the MDP [0180] F Force [0181] a Acceleration; [0182] a.sub.in Input acceleration and deceleration due to the engine and brake, respectively, according to the motion planning [0183] v Velocity [0184] v.sub.rel Relative velocity [0185] v.sub.ego, v.sub.E Velocity of ego vehicle [0186] v.sub.target, v.sub.K Velocity of target vehicle sDisplacement; [0187] E.sub.kin Kinetic energy [0188] p Momentum; [0189] ΔT Deformation energy [0190] s Displacement on a trajectory; State in the MDP s∈; empirical standard derivation [0191] {, , , } 5-tupel which defines the Markov Decision Process (MDP) with the set of states , the set of actions , the set of transitions , set of rewards e, and discount factor γ [0192] .sub.s Set of available actions a in state s action s′ Future state in the MDP s′∈ with reference to state s [0193] T(s, a, s′) Transition in the MDP between the state s and s′ according to the action a [0194] Re(s, a, s′) Reward in the MDP between the state s and s′ according to the action a [0195] f(s, a, s′) Feature in the MDP which is derived between the state s and s′ according to the action a [0196] θ Weight of the reward function [0197] π Policy (sample of II) [0198] π* Optimal policy [0199] π.sup.s Selected policy [0200] Π Set of possible policies [0201] Δt.sub.E,dyn Step size to integrate the dynamic model [0202] τ.sub.E,dyn (Δt.sub.E,dyn=t.sub.m+1−t.sub.m) V(s)Value in the MDP of the state s [0203] TH Time horizon of the planning process [0204] Ψ∈ψ Accident severity as part of risk [0205] R (criticality feature) [0206] R,.sub.f Risk (criticality feature)—in general terms and as feature in the motion planning (partly aggregated) [0207] Δv velocity change during crash/technical accident severity [0208] P(C) Collision probability [0209] Ψ.sub.impact Severity in the moment of impact [0210] R.sub.thr Risk threshold for the graceful degradation [0211] M Environment model [0212] η Accident severity prediction function [0213] ω Label for the instantaneous vehicle object (collision) configuration (ω:=z.sub.EK); Yaw rate [0214] ω.sub.coll, ω.sub.coll Instantaneous vehicle object configuration which is in collision or not in collision [0215] ω.sub.coll.sup.Pmax Most probable collision configuration [0216] Ω SZ Sample space; Random variable to the yaw rate [0217] ωΩ.sup.dyn Sample space of dynamic elements [0218] Ω.sup.stat Sample space of static elements [0219] Ω.sup.voc Sample space of vehicle object constellations [0220] τ Dynamic model (e.g., τ.sub.E,dyn represents the ego vehicle [0221] dynamic with a non-linear single track model) [0222] x Position in x direction [0223] y Position in y direction [0224] φ Orientation [0225] o.sub.k∈ The sample target vehicles o.sub.k are summarized by the set h.sub.k.sup.j∈.sub.k The sample intentions h.sub.k.sup.j of target vehicle o.sub.k are summarized by the set set .sub.k of the target vehicle [0226] EK Bidirectional event between ego vehicle E and one target object [0227] o.sub.k=K It relates to the substate z.sub.EK of one time step t.sub.i of object o.sub.k [0228] with intention h.sub.j and thus is equal to EKEK.sub.i,k.sup.j [0229] C Event collision [0230] z∈Z State vector as sample vector and random vector(e.g., z.sub.E∈Z.sub.E, z.sub.K∈Z.sub.K) [0231] x.sub.c∈X.sub.c Object classification as sample and random variable [0232] z.sub.EK∈Z.sub.EK State vector of bidirectional substate between ego E and one target object K as sample vector and random vector with [0233] EKEK.sub.i,k.sup.jz.sub.dyn∈Z.sub.dyn State vector with dynamic, time variant elements of z as sample vector and random vector [0234] z.sub.stat∈Z.sub.stat State vector with static, time invariant elements of z as sample vector and random vector [0235] g∈G State vector with geometric elements of z as sample vector and random vector [0236] {tilde over (g)}∈{tilde over (G)} Reduced state vector with geometric elements of z as sample vector and random vector [0237] p∈P Reduced state vector excluding geometric elements of z as sample vector and random vector [0238] z.sub.input∈Z.sub.input Reduced state vector with directly and indirectly measureable elements of z as sample vector and random vector [0239] Z.sub.train∈Z.sub.train Reduced state vector with non-measureable elements of z as sample vector and random vector [0240] f.sub.crit(s, a, s′) Feature for the criticality estimation [0241] f.sub.comfort(s, a, s′) Feature for the comfort estimation [0242] f.sub.progress(s, a, s′) Feature for the progress estimation [0243] ψ.sub.max Accident severity at the most probable collision configuration ω.sub.coll.sup.Pmax [0244] D, D′ etc. Minkowski Difference. The apostrophe indicates a transformed Minkowski Difference. [0245] δ Confidence interval for the angle probability; Steering angle [0246] Λ Matrix of eigenvalues [0247] V Matrix of eigenvectors [0248] α Rotation angle of eigenvectors; Angle to COG line; Accuracy function α(⋅); Slip angle [0249] Σ Covariance matrix [0250] l Length of a vehicle [0251] w Width of a vehicle [0252] e Restitution coefficient [0253] M Mass matrix of multi-body system mMass; Index of time step t.sub.m for the integration of a dynamic model [0254] C Damper matrix of multi-body system cDamper coefficient of a damper [0255] K Stiffness matrix of multi-body system [0256] k Stiffness of a spring, Index of objects o.sub.k [0257] ξ Additional static feature [0258] J Moment of inertia; [0259] f.sub.model Established models (e.g., dynamic model τ) to map direct measurable parameters [0260] ϑ Correlation coefficient [0261] q Weighting factor between self and target protection [0262] S Momentum (impact drive) [0263] B Impact point [0264] μ Friction; Expected value [0265] r Distance vector [0266] r.sub.F Distance vector to force insertion [0267] e Error (defined as difference between the estimated and reference value such as e=({circumflex over (⋅)})−(⋅); Restitution coefficient [0268] α Rotation angle of eigenvectors; angle to COG line; accuracy function α(⋅) [0269] ϕ(⋅) Relevance function [0270] F1-score harmonic mean (e.g., between precision and recall) [0271] g t) Execution time value [0272] T-score harmonic mean (e.g., between F1-score and g(t)) [0273] σ Standard derivation; Mechanical load [0274] ϵ Deformation/displacement [0275] z.sub.cm State to the crash motion model τ.sub.cm [0276] λ Eigenvalue; Constraint; Progress on trajectory (e.g., λ.sub.ego) [0277] k.sub.air Constant of the flow resistance [0278] k.sub.eEngine proportion relating to front and rear axle [0279] d Distance (e.g. distance between current position of the vehicle and the centerline [0280] d.sub.lat, distance to potential collision objects at the moment of appearance d.sub.appear [0281] f.sub.update Update frequency of the planning process

REFERENCE SIGN LIST

[0282] 1 Motor vehicle

[0283] 2 Assistance system

[0284] 3 Environment detection device

[0285] 4 Electronic computing device

[0286] 5 Environment

[0287] 6 Object

[0288] 7 Uncertainty

[0289] 8 Trajectory

[0290] 8a Evasive maneuver

[0291] 8b Collision maneuver

[0292] 8c Mitigation maneuver

[0293] 9 Collision probability

[0294] 10 Most probable impact constellation

[0295] 11 Accident severity

[0296] 12 Weight

[0297] 13 Past position

[0298] 14 Longitudinal acceleration device

[0299] 15 Lateral acceleration device

[0300] 16 Further object

[0301] S1 First step

[0302] S2 Second step

[0303] S3 Third step

[0304] S4 Fourth step

[0305] S5 Fifth step

[0306] S6 Sixth step

[0307] R Risk Value