DE102016203522A1 - Method and device for predicting trajectories of a motor vehicle - Google Patents

Abstract

The invention relates to methods in a motor vehicle (50) for predicting trajectories of the motor vehicle (50), comprising the steps of: detecting an environment (66) of the motor vehicle (50) by at least one environment detection device (2), querying parameters of the motor vehicle (50 ) and / or further motor vehicles (51, 52) by a controller (3), generating a set (28) of reference trajectories (27) up to a predetermined future time by a trajectory generator (4), the reference trajectories (27) based on the detected environment (66) and the queried parameters of the motor vehicle (50) are generated; Evaluating the generated reference trajectories (27) by a context evaluation device (5), each of the reference trajectories (27) being assigned an occurrence probability, and / or comparing the generated reference trajectories (27) with a current trajectory history (30) of the motor vehicle (50) a comparison device (6), each of the generated reference trajectories (27) being assigned in each case a similarity value, wherein the similarity value describes how large each match of the generated reference trajectory (27) with the current trajectory history (30); Selecting one of the reference trajectories (27) as a function of the occurrence probabilities and / or the similarity values associated with the reference trajectories (27) by a selection device (7), and outputting the selected reference trajectory (27) as the most plausible trajectory (34). Furthermore, the invention relates to an associated device.

Description

The invention relates to a method and a device for use in a motor vehicle for predicting trajectories of the motor vehicle.

Modern motor vehicles offer a variety of assistance systems that assist the driver in driving the motor vehicle. In this case, assistance systems are known which merely provide the driver with additional information about the motor vehicle and its surroundings, for example in the form of distance sensors or rear-view cameras. Furthermore, also partially automatic assistance systems are known, which remove the driver, for example, the steering when parking. The driver then only has to operate the accelerator pedal and the brake. Fully automatic assistance systems, which are capable of guiding the motor vehicle completely automatically, are also increasingly being used in various areas.

Modern driver assistance systems are able to detect other road users, to adapt the speed to their driving behavior and to perform an intervention for collision avoidance in critical situations. To further enhance the performance of these functions, there is a need for accurate motion prediction that can identify a future maneuver and predict the exact vehicle position.

From the DE 10 2008 005 310 A1 For example, there is known a method for influencing the movement of a vehicle upon premature detection of an unavoidable collision with an obstacle. In case of premature detection of an unavoidable collision with an obstacle as a function of determined environmental information, measures are taken to reduce the consequences of the accident. As part of the procedure probabilities of the vehicle for future times are calculated. It is provided that in case of premature detection of an unavoidable collision, the friction forces resulting from a longitudinal acceleration requested by the driver and a steering angle requested by the driver are distributed to optimize the motion trajectory of the vehicle as a function of a predicted crash behavior of the accident participants.

From the WO 2012/119596 A1 are known a method and a device for predictive indication of the possibility of evasion in highly dynamic situations for a vehicle, wherein by means of evaluation of data of an environment detection system signs for a timely lapping of an object on the lane of the own vehicle are detected and a probable behavior of the object in the near Future is predicated by hypotheses and a driver is predictively indicated an alternative option.

The invention is based on the technical problem of providing a method and an apparatus for predicting a trajectory of a motor vehicle in which the prediction is improved.

The technical problem is solved by a method with the features of claim 1 and a device having the features of claim 11. Advantageous embodiments of the invention will become apparent from the dependent claims.

In particular, a method is provided in a motor vehicle for predicting trajectories of the motor vehicle, comprising the following steps: detecting an environment of the motor vehicle by at least one environment detection device, querying parameters of the motor vehicle and / or other motor vehicles by a controller, generating a set from reference trajectories to a given future time by a trajectory generating means, wherein the reference trajectories are generated based on the detected environment and the queried parameters of the motor vehicle; Evaluating the generated reference trajectories by a context evaluation device, each of the reference trajectories being assigned an occurrence probability, and / or comparing the generated reference trajectories with a current trajectory history of the motor vehicle by a comparison device, wherein each of the generated reference trajectories is each assigned a similarity value, the similarity value describing how large a match of the generated reference trajectory with the current Trajektorienhistorie is; Selecting one of the reference trajectories as a function of the occurrence probabilities and / or the similarity values assigned to the reference trajectories by a selection device, and outputting the selected reference trajectory as the most plausible trajectory.

Furthermore, an apparatus for a motor vehicle for predicting trajectories of the motor vehicle is provided, comprising at least one surroundings detection device for detecting an environment of the motor vehicle, a control for interrogating parameters of the motor vehicle, a trajectory generation device for generating a set of reference trajectories up to a predefined future point in time, wherein the reference trajectories are generated based on the detected environment and the queried parameters of the motor vehicle; a Context evaluation device, wherein the context evaluation device is designed to evaluate the generated reference trajectories and each assign a probability of occurrence of each of the reference trajectories, and / or a comparison device which is adapted to compare the generated reference trajectories with a current Trajektorienhistorie of the motor vehicle and each of the generated reference trajectories each assign a similarity value, wherein the similarity value describes how large each match the generated reference trajectory with the current Trajektorienhistorie; and a selection device which is designed to select one of the reference trajectories as a function of the occurrence probabilities and / or the associated similarity values and to output the selected reference trajectory as the most plausible trajectory.

Parameters of the motor vehicle are intended to denote in particular a current position, a current orientation and a current speed of the motor vehicle. Furthermore, the parameters should also include performance parameters of the motor vehicle, such as an engine power and an associated ability to accelerate or decelerate (negative acceleration). In particular, the parameters determine the amount of reference trajectories generated since the endpoints and target points of the reference trajectories are also a function of the parameters. If, for example, the speed of the motor vehicle is increased, then the amount of reference trajectories in the direction of travel can shift further forward, since more target points can be achieved by the motor vehicle for a given future point in time. Another engine power, for example, alters the amount of reference trajectories that are generated, since with a higher engine power or ability to accelerate, a larger radius of target points can be achieved.

The reference trajectories can be generated, for example, by a method which is described in Michael Düring et al., "Adaptive Cooperative Maneuver Planning Algorithm for Conflict Resolution in Diverse Traffic Situations", 2014 International Conference on Connected Vehicles & Expo (ICCVE 2014), pages 238-245, November 2014, doi: 10.1109 / ICCVE.2014.7297550 ; is described.

In one embodiment, it is provided that the generation of the set of reference trajectories only takes place for a finite selection of individual destination points from a reachability set, the reachability set comprising all those points in the surroundings which can be approached by the motor vehicle up to a predefined future point in time.

In order to determine the reachability quantity, each discrete point is determined in each case under a slightly modified assumption of the lateral or longitudinal acceleration within the maximum possible range. A detailed description of the calculation of reachability quantities is given in Michael Düring et al., "Adaptive Cooperative Maneuver Planning Algorithm for Conflict Resolution in Diverse Traffic Situations", 2014 International Conference on Connected Vehicles & Expo (ICCVE 2014), pages 238-245, November 2014, doi: 10.1109 / ICCVE.2014.7297550 , to find. The selection of target points is then generated on the basis of the calculated reachability quantity. These are the endpoints of the reference trajectories or the end of every possible action. The goal is to define a discrete but nevertheless representative selection of target points from the total reachability set.

As a result, the computational effort can be significantly reduced, although it is still possible to predict a plausible reference trajectory.

For example, a grid is first set up for the selection of the target points, the grid consisting of a defined number of points in the x and y directions. The points can be equally distributed or have a different distribution, for example, to be able to make a different discretization in unimportant areas than in relevant areas. For each grid point, the nearest point from the reachability set is then searched. If the accessibility quantity is discretized too small, it is possible that the target points lie at very irregular distances from one another or an identical point for two grid points is identified as the closest one. The advantages are that a reduction in the number of reference trajectories can be made without there being any reduction in the significance of the final result and that these reference trajectories lead to selected locations, for example the lane center.

In a further embodiment it is provided that the probability of occurrence is calculated from a static component and a dynamic component, the static component taking into account static obstacles and / or road information, for example a number, alignment or limitation of lanes, and the dynamic component a behavior of others Road users and / or traffic rules considered. In this way, all objects that are in the vicinity of the motor vehicle, and a general context of a current traffic situation, such as applicable traffic rules or other standards, non-codified conventions and Behaviors are best represented and taken into account.

In a further embodiment it is provided that in the context of the evaluation by the context evaluation device of a road at least one road area is assigned, with each associated road area a stay security is defined, and wherein each reference trajectory depending on the respective target point of the reference trajectory of the corresponding value for the stay security is assigned, and wherein the respectively assigned value is included in the static evaluation of the reference trajectory. The advantage is that in this way a prioritization of road areas is made, by means of which it can be determined or assessed how likely a stay of the motor vehicle in such a road area is. For example, if the motor vehicle is located on a country lane, the proper traffic lane of the motor vehicle has a high level of safety in residence. The lane of oncoming traffic, however, a lower security and areas to the right and left of the highway, so away from the lanes, even lower security. In general, therefore, a driver of the motor vehicle will endeavor to remain in his proper lane and after an overtaking process in which he has changed to the lane of oncoming traffic, back to this proper lane back.

In one embodiment, it is further provided that in the context of the evaluation by the context evaluation device to the respective target point of each Referenztrajektorie an orientation relative to the lane orientation of a currently traveled road is determined, wherein an evaluation of the associated reference trajectory is made depending on the respective orientation, and wherein the respective evaluation is included in the static evaluation of the reference trajectory. It is assumed that a driver of a motor vehicle will endeavor to achieve a safe state of the motor vehicle at all times. In this case, an optimal orientation is a longitudinal axis of the motor vehicle oriented parallel to the road surface, so that the evaluation is carried out accordingly:
A parallel orientation to the road is rated better than a different orientation. In the worst case, the motor vehicle is oriented at a right angle to the road or even counter to the direction of travel.

In a further embodiment, it is provided that a reference trajectory is evaluated as a function of further parameters, in particular an acceleration which a motor vehicle experiences when traversing the reference trajectory, and the resulting value is included in the static evaluation of the respective reference trajectory. It is assumed that a driver of a motor vehicle has certain preferences about which accelerations he is prepared to tolerate. In this case, accelerations in both the longitudinal and in the lateral directions of the motor vehicle can be taken into account. Depending on the preference and acceleration, which would be necessary for traversing the respective reference trajectory, the reference trajectory is evaluated accordingly. If, for example, the acceleration is below a limit value specified by the driver or generally, then the corresponding reference trajectory is rated better than when the predefined limit value is reached or exceeded.

In one embodiment, it is provided that a reference trajectory is evaluated as a function of a collision with static obstacles, with which a motor vehicle would collide on this reference trajectory, and the resulting value is included in the static evaluation of the respective reference trajectory. A static obstacle is intended to mean an object which is stationary relative to the surroundings of the motor vehicle and which would cause a change of state of the motor vehicle upon contact with the motor vehicle. These include, for example, parked vehicles on the roadside, a guardrail to delimit the roadway or even trees and stones that are located off the carriageway. For the collision check, for example, the Separating Axis Theorem (SAT) method is used, which is also used for checking collisions on trajectories Michael Düring et al., "Cooperative Maneuver Planning for Cooperative Driving", IEEE IST Magazine, 2016 , is described.

It is assumed that a driver would prefer to choose a collision-free reference trajectory. A driver would always try to pass a static obstacle as far as possible with a sufficiently large distance. Small distances are only allowed if no other alternative exists. The minimum distance of a moving motor vehicle to a static obstacle is determined by the same method as the collision check.

A defined limit value decides whether this minimum distance is greater or less than the minimum distance between the motor vehicle and the static obstacle.

For example, there are three evaluation levels: collision-free, minimum distance within the defined limit range and collision-related. Since the probability of a collision in collision-free action around Orders of magnitude lower, the score for collision-related reference trajectories should be correspondingly low compared to uncritical reference trajectories or those with a small minimum distance.

Ideally, it is envisaged that all the described evaluations, ie the evaluation of the road areas, the orientation at the destination point, the occurring acceleration and a collision with static obstacles, are jointly incorporated into the static evaluation of the reference trajectories. Depending on the configuration, a total value can be formed from the individual values. In this case, it is possible for the values to be additionally provided with individual coefficients which predetermine a weighting and / or with corresponding normalization constants. In any case, the total value represents a function of the individual evaluations.

In a further embodiment, it is provided that the dynamic component of the context evaluation takes into account a collision probability of the individual reference trajectories estimated for the motor vehicle with respective estimated and evaluated reference trajectories of other motor vehicles present in the environment. The other motor vehicles present in the environment should generally include other road users, in particular also cyclists and pedestrians.

The aim of the dynamic context evaluation is to give an evaluation for each reference trajectory with regard to other road users or dynamic, ie moving, objects which allows a statement as to whether or not it is a meaningful action option for the driver. The method is based on a human-like situation analysis in that the reference trajectories are checked for collision with the estimated trajectories of the other road users. Thus, the other road users can be included in the evaluation of the respective reference trajectories.

In particular, it is provided in one embodiment that reference trajectories are also generated and evaluated for the other motor vehicles in the environment, wherein the reference trajectories of other motor vehicles present in the environment of the motor vehicle are generated and evaluated in the same way as the reference trajectories of the motor vehicle. In this case, all method steps of the static context evaluation can be carried out for every other road user or for each dynamic object in the environment of the motor vehicle.

In a further embodiment, it is provided that the ranks are ranked according to prevailing traffic rules in the surroundings of the motor vehicle, and the reference trajectories estimated for the other motor vehicles are evaluated according to their respective rank, and the respective resulting values are included in the dynamic evaluation. This makes it possible, for example, to incorporate applicable in the current environment of the motor vehicle right of way rules in the evaluation.

Furthermore, it is also possible for the ranking to be redetermined by an optimality criterion. In this way, even a highly prized motor vehicle could adapt its behavior, if that makes sense. For example, a motor vehicle traveling on a proper lane of a highway would adjust its behavior in recognizing an oncoming other motor vehicle in the proper lane: while enjoying the right of way on the proper lane, it would, for example, brake or drive to the edge of the highway.

The described method and apparatus are not only suitable for making a precise movement prediction when a driver is driving the motor vehicle, but also for providing a trajectory for fully automated driving of the motor vehicle.

Parts of the device can also be designed, individually or combined, as a combination of hardware and software, for example as program code that is executed on a microcontroller or microprocessor.

The invention will be explained in more detail with reference to preferred embodiments with reference to the figures. Hereby show:

1 a schematic representation of an embodiment of a device for a motor vehicle for predicting trajectories of the motor vehicle;

2 a simplified schematic flow diagram of the method for predicting trajectories of the motor vehicle;

3 a schematic representation of a traffic situation to illustrate the process (initial situation);

4 a schematic representation of the traffic situation to illustrate the generation of the reachability amount;

5 a schematic representation of the traffic situation to illustrate the determination of target points from the reachability amount;

6 a schematic representation of the traffic situation to illustrate the generation of the reference trajectories / options for action to target points;

7 a schematic representation of the traffic situation to illustrate the static context evaluation over road areas;

8th a schematic representation of the traffic situation to illustrate the static context evaluation on an orientation at the destination point;

9 a schematic representation of the traffic situation to illustrate the static context evaluation on an acceleration;

10 a schematic representation of the traffic situation to illustrate the static context evaluation on a collision with obstacles;

11 a schematic representation of an overall result of the static context evaluation;

12 a schematic overview diagram to illustrate the dynamic context evaluation;

13 a schematic representation of the traffic situation to illustrate the dynamic context evaluation;

14 a schematic representation of an overall result of the context evaluation;

15 a schematic representation of the traffic situation to illustrate the similarity analysis;

16 a schematic representation of the evaluation results of the similarity analysis;

17 a schematic representation of the traffic situation with end result and most plausible reference trajectory.

1 shows a schematic representation of an embodiment of a device 1 for a motor vehicle 50 for predicting trajectories of the motor vehicle 50 , The device 1 includes an environment detection device 2 , a controller 3 , a trajectory generation device 4 , a context evaluation facility 5 , a comparison device 6 and a selector 7 , The environment detection device 2 , the trajectory generator 4 , the context evaluation facility 5 , the comparison device 6 and the selector 7 are each with the controller 3 connected. Furthermore, the controller includes 3 an interface 8th about which the controller 3 with a bus system 9 of the motor vehicle 50 , For example, a CAN bus, communicate with other devices of the motor vehicle 50 Can exchange data.

The environment detection device 2 detects a current environment of the motor vehicle 50 and detects static and dynamic objects in it. The from the surroundings detection device 2 Data generated from this is the control 3 fed. The control 3 asks for parameters of the motor vehicle 50 over the interface 8th from. The parameters include in particular a current position, a current orientation and a current speed of the motor vehicle 50 , Furthermore, the parameters include performance indicators of the motor vehicle 50 such as an ability to accelerate, etc. It can be provided that the performance parameters remain constant only once when the device is used for the very first time 1 queried or entered in any other way and then in the controller 3 be kept for future use.

The data about the environment and the parameters are provided by the controller 3 to the trajectory generation device 4 forwarded. From these, it generates a set of reference trajectories up to a given future point in time and sends the generated set of reference trajectories via the controller 3 to the context evaluation facility 5 and the comparator 6 ,

The context evaluation facility 5 evaluates the individual reference trajectories as part of a static and a dynamic context evaluation and assigns values of the evaluation results corresponding to the individual reference trajectories in the form of a probability of occurrence and forwards them via the controller 3 the selection device 7 to.

Alternatively or additionally, the comparison device compares 6 the generated reference trajectories with a current Trajektorienhistorie the motor vehicle 50 and each assigns a similarity value to each of the reference trajectories, wherein the similarity value is a measure of a correspondence of a reference trajectory with the current trajectory history of the motor vehicle 50 is. The similarity values are used by the comparator 6 about the controller 4 the selection device 7 fed.

The current trajectory history 30 (please refer 15 ) is, for example, from the controller via the interface in other devices of the motor vehicle 50 queried or in the control itself by continuously storing the queried parameters of the motor vehicle 50 generated.

The selection device 7 selects one of the evaluated reference trajectories as a function of the occurrence probabilities and / or the associated similarity values. Which of the three options is chosen depends, for example, on whether a trajectory history is available or not. The selected reference trajectory is selected by the selector 7 the controller 4 supplied, which outputs this for further processing, for example via the interface 8th , In this way, other facilities of the motor vehicle 50 access the predicted trajectory. Such another device may for example be an autopilot 10 who is the motor vehicle 50 automated drives, or a security assistant 11 who intervenes only in safety-critical situations.

In 2 Figure 3 is a simplified schematic flow diagram of the method for predicting trajectories of the motor vehicle 50 shown. In the first process step 100 action possibilities are generated in the form of a set of reference trajectories by the trajectory generation device. The generated reference trajectories will be in the next step 102 evaluated by the context evaluator.

The evaluation is divided into a procedural step 103 for the static context evaluation, in which the reference trajectories are evaluated in terms of static obstacles and road information, and a procedural step 104 for the dynamic context evaluation, in which the generated reference trajectories are evaluated in the context of the behavior of other road users, taking into account the prevailing traffic rules. As part of the dynamic context evaluation, options for action of the other road users are also analyzed and evaluated by means of a static context assessment.

Alternatively or additionally, in a parallel process step 105 a similarity analysis between the respectively generated reference trajectories and a Trajektorienhistorie the motor vehicle instead, each of the reference trajectories depending on a similarity with the Trajektorienhistorie a value is assigned. In the last procedural step 106 The most plausible reference trajectory is then selected by the selection device on the basis of the results of the context evaluation and / or the similarity analysis and output as a signal or in the form of data.

In 3 is a schematic representation of a typical traffic situation 55 on a country road 60 to illustrate the method. It shows 3 the following starting situation: a motor vehicle 50 is located on a country road 60 during an overtaking process next to a truck 52 on an opposite lane 62 the highway 60 (Here, the situation for a right-hand traffic is generally shown, the same applies analogously for left-hand traffic). The truck 52 is on the proper lane 61 of the motor vehicle 50 , On the opposite lane 62 there is another motor vehicle 51 , which at a speed on the motor vehicle 50 zufährt. Both the motor vehicle 50 as well as the truck 52 indicate a speed in the direction of the proper lane 62 on.

4 shows a schematic representation of the typical traffic situation 55 on the highway 60 to clarify the process. For simplicity, only the motor vehicle is 50 shown. The same reference numerals always correspond to the same features or terms.

First, the trajectory generator is based on a current position 20 , Orientation and speed of the motor vehicle 50 an availability quantity 21 generated.

These are a discrete number of points representing all positions up to a given future time from the current position 20 of the motor vehicle 50 can be reached from. For the determination of the availability amount 21 every discrete point 22 each determined under a slightly different assumption of lateral or longitudinal acceleration within the maximum possible range. The range of values for the accelerations results from parameters (eg performance indicators) of the motor vehicle 50 , A detailed description of the calculation of availability quantities 21 is in Michael Düring et al., "Adaptive Cooperative Maneuver Planning Algorithm for Conflict Resolution in Diverse Traffic Situations", 2014 International Conference on Connected Vehicles & Expo (ICCVE 2014), pages 238-245, November 2014, doi: 10.1109 / ICCVE.2014.7297550 , to find. All other positions that are out of reachability 21 are due on the one hand because of the limited performance of the motor vehicle 50 and on the other hand for physical driving reasons (extreme braking, steering, etc.) unreachable.

In 5 is a schematic representation of the traffic situation 55 on the highway 60 to illustrate the generation of the reference trajectories. Based on the availability amount 21 become target points 23 generated. These are points at the end of every possible action. The goal is to get out of the total reachability 21 a discrete, yet representative selection of target points 23 define. It is deliberately not excluded that the driver of the highway 60 with the motor vehicle 50 can leave because such actions are quite possible in response to critical situations. However, it is unlikely that a driver deliberately tries to reach points that are particularly far off the highway 60 lie. For that reason, it does not make sense to have the complete reachability set 21 as possible destinations 23 for reference trajectories. Using a lateral constraint will become target points 23 only up to a limited distance 63 to the roadside 64 the highway 60 generated. By reducing the number of target points 23 the computational effort is significantly reduced.

For the selection of the target points 23 becomes a grid first 24 set up from a defined number of points in the x direction 25 and y direction 26 consists. To every grid point of the grid 24 then becomes the nearest point from the reachability set 21 searched. If the discretization of the availability amount is too low 21 it is possible that the target points are at very irregular intervals to each other or an identical point for two grid points is identified as the nearest point.

The procedure must therefore ensure that every destination point 23 only occurs once. In 5 is only the result of generating target points 23 based on the availability quantity 21 and the grid points of the grid 24 shown.

In 6 is a schematic representation of the traffic situation 55 on the highway 60 to illustrate the generation of the reference trajectories.

To every destination 23 In this method step, a trajectory is generated. These trajectories are the reference trajectories 27 which together the amount 28 represent all options for action. The method for generating the reference trajectories 27 based for example on the in Michael Düring et al., "Adaptive Cooperative Maneuver Planning Algorithm for Conflict Resolution in Diverse Traffic Situations", 2014 International Conference on Connected Vehicles & Expo (ICCVE 2014), pages 238-245, November 2014, doi: 10.1109 / ICCVE.2014.7297550 , described methods. The generation of the reference trajectories 27 but can be done differently. However, each of these approaches aims to produce a single optimal trajectory. The method described here, however, allows a large amount 28 different reference trajectories 27 to generate with little computational effort.

The result is a lot 28 drivable reference trajectories 27 that represent all possible options for action. The drivability check is performed by means of a vehicle model, eg the Kamm-Kreis model or a single-track model. The reference trajectories generated in this way 27 are exemplary in 6 shown.

The task is then to identify from the set of possible actions the one that the future movement of the motor vehicle 50 most likely describes.

In 7 becomes the traffic situation 55 considered for illustration in a modified form. Only the motor vehicle is used for the static context evaluation 50 in the context of a static environment 66 considered; all other road users are not taken into account. It is still a two-lane highway 60 on which the motor vehicle 50 in the same state on the opposite lane 62 moves. However, there are no other motor vehicles that influence the driving behavior of the motor vehicle 50 could have, in the environment 66 of the motor vehicle 50 ,

• 1. Position des Kraftfahrzeugs 50 innerhalb definierter Straßenbereiche 65,
• 2. Orientierung des Kraftfahrzeugs 50 am Zielpunkt 23,
• 3. longitudinale Beschleunigung,
• 4. Kollision mit möglichen statischen Hindernissen.

To a reference trajectory 27 to identify which is the most likely future course of motion of the motor vehicle 50 describes are the reference trajectories 27 evaluated in the context of a static context assessment by the Context Assessment Facility according to these four following criteria:

• 1st position of the motor vehicle 50 within defined street areas 65 .
• 2. Orientation of the motor vehicle 50 at the destination point 23 .
• 3. longitudinal acceleration,
• 4. Collision with possible static obstacles.

The car 50 is on the opposite lane 62 and has no safe place to stay. Therefore, the driver of the motor vehicle 50 with high probability try to get on the proper lane as soon as possible 61 to return. As a less unlikely can the opposite lane 62 as the future residence of the motor vehicle 50 to be viewed as. A third possibility exists in it, the lanes 61 . 62 and with it the highway 60 to leave. However, it can be assumed that the driver chooses this option only if there are no other alternatives available and if he is the area 67 off the highway 60 classifiable as passable. On the basis of in 7 illustrated traffic situation 55 can be three different areas 67 . 68 . 69 identify: the proper lane 61 , the opposite lane 62 and the area 67 outside the paved highway 60 ,

For every area 67 . 68 . 69 Each is assigned a different rating w _{stat, street} , which is a measure of the security of residence within the appropriate area 67 . 68 . 69 serves. reference trajectories 27 with destination points on the proper lane 61 are thus rated best on the opposite lane 62 however, a little lower. Is the target point of a reference trajectory 27 off the highway 60 , is used for the reference trajectory 27 given a significantly lower rating than for those reference trajectories 27 on the highway 60 lie. The exact value depends on the available information about the condition of the roadside.

7 illustrates both the division of the areas 67 . 68 . 69 as well as the ratings of the reference trajectories 27 according to the positions of their target points. The different line types represent the three evaluation levels, with a solid line indicating the best and a dotted line the lowest rating level.

The values 1, 0, 9 and 0.3 for the assessment w _{stat, road} are chosen by way of example and should be determined by means of empirical values or statistical evaluations, eg from trial studies.

In 8th is a schematic representation of the traffic situation 55 on the highway 60 to illustrate a static context evaluation on an orientation of the motor vehicle 50 at the end of a reference trajectory 27 shown.

Within this evaluation criterion becomes the respective destination point 23 a reference trajectory 27 after orientation 56 of the motor vehicle 50 in terms of road orientation 70 the highway 60 rated. A small angle in this context means a parallel position of the vehicle to the road. By contrast, a large angle expresses that the longitudinal axis of the motor vehicle 50 opposite the course of the highway 60 or the lanes 61 . 62 is twisted. This corresponds to the orientation 56 another safety criterion. An important variable for determining orientation 56 of the motor vehicle 50 is the vehicle orientation, which is the orientation of the motor vehicle 50 , for example in a global coordinate system. To calculate the orientation 56 the difference between the vehicle orientation and the angle must be formed, which is the road orientation 70 the highway 60 or the lanes 61 . 62 in the global coordinate system. In a special case, that the country road 60 parallel to the global x-axis, the orientation is the same 56 exactly the road orientation 70 the highway 60 ,

The rating according to the orientation 56 of the motor vehicle 50 at the respective destination point 23 a reference trajectory 27 assumes that every driver has the intention to be in a safe state at all times during his journey. An orientation 56 of 0 ° corresponds to the ideal state of the motor vehicle 50 with an optimal location on a lane 61 . 62 , For angles with a value significantly greater than 0 ° there is a risk that the motor vehicle 50 after passing through the reference trajectory 27 the roadway 61 . 62 leaves, as a steering correction is then no longer possible.

In 8th are the ratings of the reference trajectories 27 characterized by the resulting three evaluation levels by means of different lines. The solid lines are reference trajectories 27 with secure destinations 23 whereas the dotted lines mark those reference trajectories 27 with large angles to the roadway 61 . 62 , A dashed line marks a value between the two extremes. The three cases are assigned, for example, the following values for the evaluation w _{stat, orientation} : 1 for an angle less than 5 °, 0.8 for an angle between 5 ° and 10 ° and 0.6 for an angle greater than 10 °. The values are chosen only by way of example and can also be dimensioned differently, for example, individual preferences, properties of the motor vehicle 50 or a road condition etc.

In 9 is a schematic representation of the traffic situation 55 on the highway 60 to illustrate further static context evaluations, for example via an assessment of an acceleration of the motor vehicle 50 when passing through a reference trajectory 27 , shown. Here, by way of example only the longitudinal acceleration of the motor vehicle 50 , that is, the acceleration along a longitudinal axis of the motor vehicle 50 , considered. In principle, however, other directions in which an acceleration can take place are also taken into account, for example accelerations in lateral directions.

With the longitudinal acceleration, a comfort aspect flows into the evaluation of the reference trajectories 27 with a. This should contribute to a set of equally rated reference trajectories 27 to identify the most likely course of action. The evaluation after the longitudinal acceleration of the motor vehicle 50 is based on the thesis that every driver is interested in driving as comfortably as possible and maintaining his current movement. Deviating intentions of the driver, eg a targeted acceleration, could also be considered accordingly, for example if an overtaking maneuver is being carried out and a greater acceleration is accepted than usual.

Three evaluation levels are defined, each of which is based on defined limit values. reference trajectories 27 with accelerations between 0 and the first limit fall into the top rank, acceleration values greater than the first, but less than a second threshold, are assigned to the second range, and at accelerations greater than an upper limit are assigned the lowest score. Analogously, the classification of the evaluation stages for a negative acceleration (deceleration) takes place, whereby high delays are rated worse than low ones. The ratings of the individual stages are identical to those of the three acceleration stages.

9 illustrates the classification of the individual evaluation levels for an amount of longitudinal acceleration. Here, the first limit at 2 m / s ^{2 is} selected, the second limit at 5 m / s ² . The three areas are each assigned the values w _{stat, acceleration} of 1, 0.95 and 0.9. The values of 2 m / s ² and 5 m / s ² are examples and can be determined, for example, by empirical values or subject studies. The reference trajectories evaluated according to the three ranges of acceleration 27 are marked by the nature of the line, with a solid line marking the first area, a dashed line the second area, and a dotted line the third area.

In 10 is a schematic representation of the traffic situation 55 on the highway 60 to illustrate a further static context evaluation via an assessment of possible collisions of the motor vehicle 50 with static obstacles when traversing a reference trajectory 27 shown.

Static obstacles summarize all immobile objects that cause a change of state of the vehicle upon contact. These include, for example, parked vehicles on the roadside, a guardrail to delimit the roadway or even trees and stones that are located off the carriageway. For the collision check of entire trajectories, for example, the Separating Axis Theorem (SAT) method is used, which in Michael Düring et al., "Cooperative Maneuver Planning for Cooperative Driving", IEEE IST Magazine, 2016 , described for this case.

It is assumed that the driver in the traffic situation 55 preferably collision-free reference trajectories 27 in comparison to collision-related reference trajectories 27 chooses. Furthermore, a driver will always try a static obstacle 71 if possible with a sufficiently large distance to pass. Small distances are only allowed if no other alternative exists. The minimum distance of a moving motor vehicle 50 to a static obstacle 71 is determined by the same method as the collision check. A defined limit value is used to decide whether this minimum distance is greater or less than the minimum distance between the motor vehicle 50 and the static obstacle 71 is.

As a result, there are three evaluation levels for the reference trajectories 27 : collision-free, minimum distance within the defined limit and collision-prone. Because the probability of colliding with a static obstacle 71 in collision-free reference trajectories 27 order of magnitude lower, the score should be for collision-biased reference trajectories 27 be correspondingly low compared to collision-free reference trajectories 27 and those reference trajectories 27 with a small minimum distance. Example values for the individual reference trajectories 27 are in 10 in which only collisions with static obstacles, but not the agreement from the road, are assessed. In this example, a minimum distance of 0.7 m was chosen. The values for the rating w _{stat, collision} are determined by the type of line of the reference trajectory 27 marked. A solid line means a value of 1, a dashed line a value of 0.8, and a dotted line a value of 0.01.

11 shows a schematic representation of an overall result of the static context evaluation, in which the four partial results w _{stat, road} , w _{stat, orientation} , w _{stat, acceleration} and w _{stat, collision are shown} together in a final result. The final result is a function of the individual evaluation variables w _{stat, road} , w _{stat, orientation} , w _{stat, acceleration} and w _{stat, collision} . In the simplest case, a weighted or unweighted sum of these four valuation variables is formed. The function can in principle be designed differently.

In 11 are the different values for the final results of the reference trajectories 27 marked by the type of line. A solid line means a large probability of occurrence w _stat , a dashed line means an average probability of occurrence w _stat and a dotted line means a low probability of occurrence w _stat . It can be seen that the static context evaluation reveals that one driver is returning from the opposite lane 62 on the proper lane 61 would prefer.

Previously, the options for action were based solely on the available information about the stationary objects, ie the roadway 61 . 62 and static obstacles 71 etc., in the environment of the motor vehicle 50 rated. Other road users were ignored. The goal of the dynamic context evaluation by the context evaluator is to, for each reference trajectory 27 to make an assessment with respect to other road users, such a rating then makes it possible to say whether it is a meaningful action option for the driver or not. The procedure is based on a human-like situation analysis by the reference trajectories 27 be checked for collision with the estimated trajectories of the other road users. The complete functional principle of the dynamic context evaluation is shown schematically in 12 shown.

This in 12 The diagram shown on the previous example scenario, in which for the motor vehicle 50 a trajectory should be predicted and in the traffic situation 55 both the truck 52 as well as the further motor vehicle 51 representing the other road users. The actual evaluation process of the possibilities of action of the motor vehicle 50 takes place only in the last, step G of the procedure. Here are the reference trajectories 27 checked for collision with the estimated trajectories of other road users and provided with appropriate ratings. In order to be able to carry out a collision check, however, the knowledge about the future trajectories of the other road users is necessary.

First, in each case a first method step A1, B1, C1 for all three motor vehicles 50 . 51 . 52 generates a set of reference trajectories. The generation of these reference trajectories takes place according to the same principle as previously used for generating the options of action of the motor vehicle 50 was applied. Within the static context evaluation A2, B2, C2, which is likewise carried out according to the method steps already described, a ranking of the created reference trajectories results, which depend on the amount of their evaluations. In this way, reference trajectories classified as probable courses of action based on the static context evaluation rank first, while unlikely reference trajectories move to positions below.

To establish the final reference trajectories for each of the three motor vehicles 50 . 51 . 52 Individual priorities, which are determined by a prioritization D, are used. An assignment of ranks indicates which position in the ranking the respective motor vehicle has regarding its right of way. The car 50 . 51 . 52 Consequently, the highest priority has over the lower priority motor vehicles 50 . 51 . 52 from an objective point of view always has priority in the choice of its course of motion or the corresponding reference trajectory. Within a dynamic analysis E, this precedence principle is adopted by determining the final reference trajectories in the order of prioritization.

The dynamic analysis E completes the prediction for the other road users. As a result, considering the F reference trajectories of the motor vehicle 50 in the last method step G, two trajectories are output, each of which is the estimated course of movement of the truck 52 and the further motor vehicle 51 next to the motor vehicle 50 describe.

In the exemplary traffic situation 55 Thus, the trajectories of the truck can be 52 and the further motor vehicle 51 determine and include in the context evaluation. The result of the dynamic context evaluation is finally a _{weighting variable} w _dyn , which for each reference trajectory of the motor vehicle 50 contains a rating depending on the collision check.

13 illustrates the dynamic evaluation of each reference trajectory 27 , The type of line again marks the value of the dynamic rating. A solid line marks collision-free action options and a dotted line of collision-related options for action. reference trajectories 27 , which are shown by a dashed line, have a minimum distance between the vehicles 50 . 51 . 52 on.

So far, only information from the environment was used to determine the most likely behavior of the predicted motor vehicle 50 close. It is possible, exclusively with the help of the evaluations from the context analysis, one reference trajectory 27 predict. The solution to this context-based prediction is determined by the combination of static and dynamic score results. Such a solution contains for each reference trajectory 27 of the motor vehicle 50 an assessment by means of which it is possible to make a statement about the plausibility of the corresponding option based on contextual knowledge.

In the 11 and 12 It has already been shown by way of example how the individual results from the static and the dynamic evaluation could look like. In 14 Now, the combination of the two sub-results is shown as a joint context evaluation. The type of line marks the value of the final result of the joint context evaluation. A solid line indicates a high probability or plausibility that a driver has the corresponding reference trajectory 27 would choose. A dashed line marks an average probability or plausibility and a dotted line marks a low probability or plausibility for the corresponding reference trajectory 27 ,

The future movement of the motor vehicle 50 results as a logical consequence of the interaction of all environmental parameters. The resulting solution describes from an objective perspective which of the options for action is the most plausible and therefore safest option for the occupants of the motor vehicle 50 represents.

The prediction is subject to the assumption that every driver in the present situation for logical reasons would choose this option as a future maneuver. In reality, however, due to human misperceptions or individual preferences and abilities, there may be strong deviations between the logical and actual course of the movement. By way of example, this problem becomes even clearer: An overtaking maneuver is predicted in which the motor vehicle 50 However, only small distances to the other road users 51 . 52 having. The driver is uncertain because of these small tolerances, whether he can implement the vehicle movement exactly according to this requirement and decides instead for a strong braking maneuver.

• 1. Der Fahrer muss die Handlungsmöglichkeit als potentielle Lösung erkennen,
• 2. der Fahrer muss die Handlungsmöglichkeit ebenfalls als beste Lösung bewerten,
• 3. der Fahrer muss sich selbst zutrauen, die Handlung auszuführen,
• 4. die tatsächlich ausgeführte Bewegung darf nicht weit von der vorhergesagten Trajektorie abweichen

There are exactly four prerequisites must be met, so that a predicted course of action or reference trajectory 27 the actual vehicle movement corresponds to:

• 1. The driver must recognize the possibility of action as a potential solution
• 2. the driver must also rate the option as the best solution,
• 3. the driver must trust himself to carry out the action
• 4. The actual movement must not deviate too far from the predicted trajectory

For this reason, as a supplement to the context evaluation, a similarity analysis is performed, which additionally includes a current trajectory history 30 considered. Nonetheless, the key advantage that logical prediction provides on the basis of contextual information is its usability in situations where there is no or very limited current trajectory history for an environment 30 of the motor vehicle 50 is available.

In 15 a further method step is shown, which schematically shows the similarity analysis (see overview diagram in FIG 2 ) illustrated by the comparison device of the device. Through the similarity analysis, it is possible the current trajectory history 30 So one of the motor vehicle 50 already trajectory, which up to the current position 20 of the motor vehicle 50 is enough to consider in the prediction.

The similarity analysis can be understood as a verification or as an extension of the result of the context evaluation. However, it is also possible to make an independent prediction based solely on the similarity analysis.

This case occurs, for example, if no or only insufficient information about the environment of the motor vehicle 50 be available. The principle of the similarity analysis is a current trajectory history 30 on similarity, ie on agreement, with the established reference trajectories 27 to investigate.

The basis for determining the similarity between the reference trajectories 27 and the current trajectory history 30 , which corresponds to the previous driving behavior, is an existing trajectory that depicts a real behavior of the past. Characteristic for the description of an associated real trajectory their observation time t is _obs * (in the example described was selected for s a value of 0.7). This indicates the period over which data about the real behavior of the past has been recorded. Crucial here is that the start time 31 the observation of the current trajectory history 30 a starting point 32 the respective reference trajectory to be evaluated 27 equivalent. Otherwise, the data must be adjusted accordingly, for example by only a section of the current trajectory history 30 is looked at. The start time 31 can do it rolling with the prediction the reference trajectory 27 for the motor vehicle 50 be moved, so that for a current time always a current Trajektorienhistorie 30 can be provided. The current trajectory history 30 then always describes a certain section of the past, which shifts with the passage of time as in a shift register into the present.

To check the similarity are cutouts 33 the reference trajectories 27 each with the current trajectory history 30 compared. The length of the cutouts 33 must match the length of the current trajectory history 30 to match. For every reference trajectory 27 Thus, a similarity value can be determined, which indicates how high the agreement between the respective reference trajectory 27 and the current trajectory history 30 is. The result is a variable w _sim with the corresponding similarity values for each reference trajectory 27 , The reference trajectory 27 with the greatest similarity is finally the predicted course of action.

16 shows all reference trajectories 27 , where the similarity values are marked by the type of line. Here, a solid line means a great similarity, a dashed line a middle similarity and a dotted line a low similarity of the reference trajectory 27 with the current trajectory history 30 ,

The method for predicting trajectories of the motor vehicle 50 is based on the two evaluation procedures Contextual Assessment and Similarity Analysis. With each of the two evaluation methods, an independent prediction is possible, independent of the other evaluation method. However, both approaches may have individual weaknesses. A possible weakness of the contextual assessment is, for example, a lack of reference to the reality of human behavior, given a current trajectory history 30 disregarded. The limit of the similarity analysis, on the other hand, is the need to continuously record a trajectory history, which however may not always be available (at the start of the journey, etc.).

By combining both approaches, however, it is possible to override the limits of the process by exploiting the respective strengths. The result is an improved method for predicting a trajectory of the motor vehicle 50 that can predict future behavior based on both real vehicle data from the past and contextual information.

The combination of the three evaluation variables from static and dynamic context evaluation and the similarity analysis can be expressed, for example, by a plausibility value P. The plausibility value P of a single reference trajectory indicates how plausible, ie how probable, for this reference trajectory 27 belonging action option as a solution for the future movement of the motor vehicle 50 is estimated. The plausibility value P is a function of the individual evaluation variables w _stat , w _dyn and w _sim . In the simplest case, a weighted or unweighted sum of these three valuation variables is formed. The function for the plausibility P can in principle be designed differently.

In 17 is the end result of the prediction based on the in the 3 to 17 described example. The reference trajectories 27 are marked P depending on their plausibility. The solid lines are reference trajectories 27 on the other hand, the dashed and dotted lines belong to reference trajectories 27 each with medium and low plausibility values. The most plausible reference trajectory predicted in the final result 34 is marked by a bold solid line. In this case, the driver becomes the roadway with the motor vehicle 50 Leave in direction of travel to the left.

The selection device then selects exactly the reference trajectory 27 with the largest plausibility value and gives this as the most plausible trajectory 34 out.

LIST OF REFERENCE NUMBERS

1

 contraption

2

 Surroundings sensing device

3

 control

4

 Trajektoriengenerierungseinrichtung

5

 Context evaluator

6

 comparator

7

 selector

8th

 interface

9

 bus system

10

 autopilot

11

 security assistant

20

 actual position

21

 reachable set

22

 discrete point

23

 Endpoint

24

 grid

25

 x-direction

26

 y-direction

27

 reference trajectory

28

 amount

30

 current trajectory history

31

 Start time

32

 starting point

33

 neckline

34

 most plausible reference trajectory

50

 motor vehicle

51

 another motor vehicle

52

 Lorry

55

 traffic situation

56

 vehicle registration

60

 country road

61

 proper lane

62

 opposite lane

63

 limited distance

64

 roadside

65

 road area

66

 environment

67

 Area outside the paved highway

68

 Area of proper lane

69

 Area of opposite lane

70

 alignment

71

 static obstacle

100-106

 steps

A1

 Process step for the motor vehicle

A2

 Process step for the motor vehicle

B1

 Process step for the truck

B2

 Process step for the truck

C1

 Process step for the further motor vehicle

C2

 Process step for the further motor vehicle

D

 prioritization

e

 dynamic analysis

F

 Consideration of reference trajectories

G

 Trajektorienausgabe

QUOTES INCLUDE IN THE DESCRIPTION

This list of the documents listed by the applicant has been generated automatically and is included solely for the better information of the reader. The list is not part of the German patent or utility model application. The DPMA assumes no liability for any errors or omissions.

Cited patent literature

• DE 102008005310 A1 [0004]
• WO 2012/119596 A1 [0005]

Cited non-patent literature

• . Michael Düring et al, "Adaptive Cooperative Maneuver Planning Algorithm for Conflict Resolution in various traffic situations," 2014 International Conference on Connected Vehicles & Expo (ICCVE 2014), pages 238-245, Nov. 2014, doi: 10.1109 / ICCVE.2014.7297550 [ 0011]
• . Michael Düring et al, "Adaptive Cooperative Maneuver Planning Algorithm for Conflict Resolution in various traffic situations," 2014 International Conference on Connected Vehicles & Expo (ICCVE 2014), pages 238-245, Nov. 2014, doi: 10.1109 / ICCVE.2014.7297550 [ 0013]
• Michael Düring et al., "Cooperative Maneuver Planning for Collaborative Driving", IEEE IST Magazine, 2016 [0020]
• . Michael Düring et al, "Adaptive Cooperative Maneuver Planning Algorithm for Conflict Resolution in various traffic situations," 2014 International Conference on Connected Vehicles & Expo (ICCVE 2014), pages 238-245, Nov. 2014, doi: 10.1109 / ICCVE.2014.7297550 [ 0063]
• . Michael Düring et al, "Adaptive Cooperative Maneuver Planning Algorithm for Conflict Resolution in various traffic situations," 2014 International Conference on Connected Vehicles & Expo (ICCVE 2014), pages 238-245, Nov. 2014, doi: 10.1109 / ICCVE.2014.7297550 [ 0068]
• Michael Düring et al., "Cooperative Maneuver Planning for Collaborative Driving," IEEE IST Magazine, 2016 [0086]

Claims (11)

Method in a motor vehicle ( 50 ) for predicting trajectories of the motor vehicle ( 50 ), comprising the steps of: detecting an environment ( 66 ) of the motor vehicle ( 50 ) by at least one environment detection device ( 2 ), Querying parameters of the motor vehicle ( 50 ) and / or other motor vehicles ( 51 . 52 ) by a controller ( 3 ), Generating a set ( 28 ) of reference trajectories ( 27 ) until a given future time by a trajectory generation device ( 4 ), the reference trajectories ( 27 ) on the basis of the recorded environment ( 66 ) and the queried parameter of the motor vehicle ( 50 ) be generated; Evaluate the generated reference trajectories ( 27 ) by a context evaluation device ( 5 ), each of the reference trajectories ( 27 ) in each case an occurrence probability is assigned, and / or comparison of the generated reference trajectories ( 27 ) with a current trajectory history ( 30 ) of the motor vehicle ( 50 ) by a comparison device ( 6 ), each of the generated reference trajectories ( 27 ) is assigned a similarity value, wherein the similarity value describes how large a respective match of the generated reference trajectory ( 27 ) with the current trajectory history ( 30 ); Selecting one of the reference trajectories ( 27 ) as a function of the occurrence probabilities and / or the reference trajectories ( 27 ) assigned similarity values by a selection device ( 7 ), and outputting the selected reference trajectory ( 27 ) as the most plausible trajectory ( 34 ). Method according to one of the preceding claims, characterized in that the generation of the quantity ( 28 ) of reference trajectories ( 27 ) only for a finite selection of individual target points ( 23 ) from a reachability quantity ( 21 ), whereby the reachability quantity ( 21 ) all those points in the environment ( 66 ), which of the motor vehicle ( 50 ) can be approached until a given future time. Method according to one of claims 1 to 2, characterized in that the probability of occurrence is calculated from a static portion and a dynamic portion, wherein the static portion static obstacles ( 71 ) and / or road information, and the dynamic component considers behavior of other road users ( 51 . 52 ) and / or traffic rules. Method according to claim 3, characterized in that in the context of the evaluation by the context evaluation device ( 5 ) of a road ( 60 ) at least one street area ( 65 ) is assigned to each assigned road area ( 65 ) a residence security is defined, and wherein each reference trajectory ( 27 ) depending on the respective target point ( 23 ) of the reference trajectory ( 27 ) the corresponding value for the residence safety is assigned, and wherein the respectively assigned value in the static evaluation of the reference trajectory ( 27 ). Method according to claim 3 or 4, characterized in that in the context of the evaluation by the context evaluation device ( 5 ) to the respective destination point ( 23 ) of each reference trajectory ( 27 ) an orientation ( 56 ) relative to the roadway orientation ( 70 ) of a currently traveled road is determined, wherein an evaluation of the associated reference trajectory is made as a function of the respective orientation, and wherein the respective evaluation is included in the static evaluation of the reference trajectory. Method according to one of claims 4 to 5, characterized in that a reference trajectory ( 27 ) as a function of further parameters, in particular an acceleration, which a motor vehicle ( 50 ) when traversing the reference trajectory ( 27 ), and the resulting value is included in the static evaluation of the respective reference trajectory ( 27 ). Method according to one of claims 3 to 6, characterized in that a reference trajectory ( 27 ) depending on a collision with static obstacles ( 71 ), by which a motor vehicle ( 50 ) on this reference trajectory ( 27 ) and the resulting value in the static evaluation of the respective reference trajectory ( 27 ). Method according to one of claims 3 to 7, characterized in that the dynamic portion of the context evaluation a probability of collision of the individual for the motor vehicle ( 50 ) estimated reference trajectories ( 27 ) with respectively estimated and evaluated reference trajectories of others in the environment ( 66 ) existing motor vehicles ( 51 . 52 ) considered. Method according to claim 8, characterized in that for the other motor vehicles ( 51 . 52 ) in the environment ( 66 ) also reference trajectories are generated and evaluated, whereby the reference trajectories of others in the environment ( 66 ) of the motor vehicle ( 50 ) existing motor vehicles ( 51 . 52 ) are generated and evaluated in the same way as the reference trajectories ( 27 ) of the motor vehicle ( 50 ). Method according to claim 8 or 9, characterized in that the other motor vehicles ( 51 . 52 ) in the environment ( 66 ) of the motor vehicle ( 50 ) Are assigned to a rank order, the ranking being determined according to prevailing traffic rules, and those for the other motor vehicles ( 51 . 52 ) are estimated according to their respective rank, and the respective resulting values are included in the dynamic rating. Contraption ( 1 ) for a motor vehicle ( 50 ) for predicting trajectories of the motor vehicle ( 50 ), comprising: at least one environment detection device ( 2 ) for detecting an environment ( 66 ) of the motor vehicle ( 50 ), a controller ( 3 ) for interrogating parameters of the motor vehicle ( 50 ), a trajectory generation device ( 4 ) for generating a quantity ( 28 ) of reference trajectories ( 27 ) up to a given future time, the reference trajectories ( 27 ) on the basis of the recorded environment ( 66 ) and the queried parameter of the motor vehicle ( 50 ) be generated; a context evaluation facility ( 5 ), the context evaluation facility ( 5 ) is designed such that the reference trajectories generated ( 27 ) and each of the reference trajectories ( 27 ) each assign a probability of occurrence, and / or a comparison device ( 6 ), which is designed such that the generated reference trajectories ( 27 ) with a current trajectory history ( 30 ) of the motor vehicle ( 50 ) and each of the generated reference trajectories ( 27 ), in each case assigning a similarity value, the similarity value describing how large a respective match of the generated reference trajectory ( 27 ) with the current trajectory history ( 30 ); and a selector ( 7 ), which is designed, one of the reference trajectories ( 27 ) depending on the occurrence probabilities and / or the associated similarity values and the selected reference trajectory ( 27 ) as the most plausible trajectory ( 34 ).

Images