Trajectory prediction method, method of controlling host vehicle, calculation unit, and program product

Abstract

The invention relates to a computer-implemented method for trajectory prediction: receiving, by a prediction module, trajectory data of a movement trajectory of a traffic participant arranged in the environment of a host vehicle, the trajectory data being arranged in a graphical representation; receiving, by the prediction module, map data depicting a map representation of an environment of the host vehicle; generating interactive graphic representation for the plurality of traffic participants through the prediction module based on the trajectory data of the traffic participants and the traffic lane pose information represented by the map; a motion trajectory to be implemented in the future is predicted for at least one further traffic participant by the prediction module on the basis of trajectory data, map data and the interactive graphical representation of the traffic participant. The invention further relates to a computer-implemented method for controlling a host vehicle, to a computing unit and to a computer program product.

Description

Technical Field

[0001] The invention relates to a method for trajectory prediction and a method for controlling a host vehicle. Background Art

[0002] Methods for trajectory prediction and methods for controlling a host vehicle are known from the prior art. Summary of the invention

[0003] It is therefore an object of the present invention to provide an improved method for trajectory prediction and an improved method for controlling a host vehicle.

[0004] This object is achieved by a method for trajectory prediction, a method for controlling a host vehicle, a computing unit and a computer program product. Advantageous embodiments are described below.

[0005] According to one aspect, a computer-implemented method for trajectory prediction is provided, the method comprising:

[0006] Receiving, by means of a prediction module, trajectory data of movement trajectories of traffic participants arranged in an environment of the host vehicle, wherein the trajectory data are arranged in a graph representation, wherein nodes of the graph representation include position information and / or speed information of traffic participants, wherein edges of the graph representation define time relationships between position information and / or speed information of traffic participants;

[0007] receiving, by means of a prediction module, map data representing a map representation of an environment of the host vehicle, wherein the map data are arranged in a graphical representation, wherein nodes of the graphical representation comprise position information of lanes and / or lane boundary elements of lanes traveled by the host vehicle and / or by other road users, wherein edges of the graphical representation define spatial relationships between the position information of lane boundary elements;

[0008] generating, by means of a prediction module, an interactive graphical representation for a plurality of traffic participants based on the trajectory data of the traffic participants and the lane pose information represented by the map, wherein each node of the interactive graphical representation shows a traffic participant located in the environment of the host vehicle and comprises position information and / or speed information of the traffic participant, wherein each edge of the interactive graphical representation defines an arrangement relationship between two of the traffic participants represented by the node;

[0009] The prediction module predicts a movement trajectory to be carried out in the future for at least one other road user based on the trajectory data, the map data and the interactive graphical representation of the road user.

[0010] The following technical advantage can thereby be achieved: an improved method for trajectory prediction can be provided. By taking into account the trajectory data and map data in a graphical representation, the trajectory data and map data can be used more simply as input data in the prediction module. In this case, the graphical representation has the advantage that it is geometrically invariant and does not require a predefined order or symmetry of the corresponding data entries. This simplifies the handling of the data.

[0011] Furthermore, by taking into account the trajectory data in conjunction with the map data, the trajectory prediction can be configured more accurately. In this case, the information of the trajectory data and the map data is incorporated in a combined manner into the prediction of the movement trajectory.

[0012] By taking into account the interactive graphic representation, the trajectory prediction can be further refined. The interactive graphic representation provides information about the relative arrangement of the individual traffic participants with respect to one another. By taking into account the relative arrangement of the traffic participants with respect to one another, the trajectory prediction can be improved as follows: the interactions of the traffic participants with respect to one another can be predicted and simultaneously taken into account.

[0013] The interaction involves an adaptation of the speed and / or a change of lane and / or lane of a traffic participant due to the relative arrangement of the other traffic participants with respect to the respective traffic participant. By taking into account possible maneuvers of the traffic participant caused by the relative arrangement of the other traffic participants, the prediction of the possible movement trajectory of the traffic participant can be refined and thus improved.

[0014] Besides this, the interactive graphical representation in the form of an interactive graphical representation enables simple handling and interaction of the data as simple use as input data for the prediction module.

[0015] According to one embodiment, the method further comprises:

[0016] An anchor path is generated by a prediction module based on trajectory data of a traffic participant and map data of a map representation, wherein the anchor path defines the following area on a lane traveled by the traffic participant: possible movement trajectories can be arranged in the area; wherein the movement data is predicted based on the trajectory data, the map data, the interactive graphical representation of the traffic participant and / or the anchor path.

[0017] The following technical advantages can be achieved in this way: by taking into account the anchoring paths, possible motion trajectories can be limited to a spatial area. Therefore, according to this embodiment, all possible motion trajectories must extend within the anchoring paths. By configuring the anchoring paths in such a way that they are limited to the direction of the corresponding lanes and / or lanes and are otherwise compatible with the current traffic regulations, possible motion trajectories can be limited to meaningful areas of the lanes and / or lanes. The anchoring paths are limited to lanes or lanes that are released by traffic participants. Anchoring paths for sidewalks and bicycle lanes are excluded from consideration. Therefore, the prediction of motion trajectories that do not follow the corresponding direction of the lanes and / or lanes and / or do not comply with the current traffic regulations can be avoided.

[0018] According to one embodiment, the generation of the anchor path includes:

[0019] By means of the prediction module, at least one traffic participant is located on the lane and / or the roadway;

[0020] recognizing a lane and / or a lane segment and / or a lane segment on which the road user is located, wherein the lane segment and / or the lane segment defines a sub-region of the lane and / or the lane, wherein the lane segment and / or the lane segment is bounded by lane boundary elements of the lane and / or lane boundary elements of the lane, wherein a lane segment can be connected to further lane segments and / or a lane segment can be connected to further lane segments, wherein a plurality of lane segments delineate a lane and / or a plurality of lane segments delineate a lane;

[0021] Taking into account the prevailing traffic regulations and the course of the lanes and / or lanes, lane segments and / or lane segments are connected to further lane segments and / or lane segments that can each be connected to one another to form an anchoring path.

[0022] The following technical advantages can be achieved in this way: An anchor path can be generated precisely. By first identifying the lane segments and / or lane segments on which the traffic participant for whom the future movement trajectory is to be predicted is situated, the corresponding anchor path is directly limited to the lane and / or lane on which the corresponding traffic participant is actually situated. This ensures that the predicted movement trajectory is directly associated with the current position of the corresponding traffic participant.

[0023] By connecting a lane segment and / or lane segment to another lane segment and / or lane segment in order to generate an anchor path based on the corresponding lane segment and / or lane segment, the corresponding anchor path can be generated in a manner extending along different lanes and / or lanes. In particular, multiple anchor paths can be generated for a traffic participant, which start from the same lane segment and / or lane segment and have different courses.

[0024] Here, different lane segments and / or lane segments are connected to form possibly multiple anchor paths in such a way that the current traffic regulations are observed. The anchor paths thus generated thus represent the following spatial regions: possible movement trajectories can be arranged in the said spatial regions.

[0025] In this way, a plurality of possible movement trajectories that can be driven by the road user according to the prevailing traffic regulations can be predicted.

[0026] According to one specific embodiment, the anchoring path includes a lane change and / or a lane change if the lane change and / or the lane change is compatible with the lane and / or the course of the lane and with the applicable traffic regulations.

[0027] The following technical advantage can be achieved in this way: a movement trajectory with lane changes and / or lane changes can be predicted. As a result, the trajectory prediction can be further refined.

[0028] According to one embodiment, the method further comprises:

[0029] By means of an environmental detection device, based on environmental sensor data of at least one vehicle, other traffic participants are classified as one of the following definitions: vehicle, bus, truck, tram, motorcyclist, cyclist, pedestrian, animal; wherein an anchor path is only generated for the following traffic participants: the traffic participants are classified as vehicle, bus, truck, tram, motorcyclist.

[0030] The following technical advantage can thereby be achieved: the generation is only carried out for those traffic participants for whom the anchor path is a meaningful limitation of the corresponding possible movement trajectory. According to the invention, the anchor path is limited to the lane and / or lane. The anchor path is therefore primarily meaningful for those traffic participants who move primarily or exclusively on the lane and / or lane. Therefore, limiting the anchor path to the vehicle prevents the generation of an anchor path for those traffic participants for whom the anchor path is not a meaningful limitation. Such traffic participants are, for example, pedestrians. As a result, the trajectory prediction is further refined.

[0031] According to one embodiment, the method further comprises:

[0032] receiving, by the prediction module, further map data which depict a map representation of an environment of the host vehicle, wherein the further map data are in the form of image data;

[0033] Lane characteristics of a lane traveled by a traffic participant are read from map data by a prediction module; wherein the movement trajectory is predicted based on trajectory data, map data and / or interactive graphic representations of traffic participants and / or lane characteristics.

[0034] This can achieve the following technical advantage: By taking into account the lane characteristics, additional information that is important for trajectory planning can be taken into account. This can further increase the accuracy of the trajectory prediction, since these lane characteristics can influence the behavior of traffic participants depending on their configuration.

[0035] In the sense of the present application, a lane characteristic is a property of a lane or lane that can influence the behavior of traffic participants traveling on the respective lane or lane. The behavior of traffic participants describes the behavior of traffic participants in road traffic and includes the execution or non-execution of maneuvers and / or the manner in which maneuvers are executed by traffic participants as actions or reactions to the prevailing traffic.

[0036] According to one embodiment, the lane features include: stop lines, traffic signs, traffic lights, pedestrian crossings, lane environments, sidewalks, and parking lots.

[0037] The following technical advantage can thus be achieved: by taking into account the lane characteristics, additional information that is important for trajectory prediction can be taken into account. All lane characteristics mentioned here can directly influence the behavior of traffic participants who pass through the lane area on which such lane characteristics are arranged. By taking into account the lane characteristics, the behavior of traffic participants and in particular the correspondingly foreseeable maneuvers of traffic participants can be included in the trajectory prediction. As a result, the trajectory prediction can be further refined.

[0038] According to one embodiment, the arrangement relationship of the edges of the interactive graphical representation includes lane information about the traffic participant, wherein the lane information includes:

[0039] The road user is located in the same lane as other road users and / or the own vehicle; and / or

[0040] The road user is located in a lane that is located next to a lane of another road user and / or of the own vehicle; and / or

[0041] The road user is located in a lane that intersects with the lane of another road user and / or the lane of the own vehicle,

[0042] The prediction of the movement trajectory is performed based on the trajectory data, the map data, the interactive graphic representation of the traffic participant and / or the lane characteristics and / or the arrangement relationship of the interactive graphic representation.

[0043] This can achieve the following technical advantage: The relative arrangement of the traffic participants can be precisely determined. By assigning different traffic participants to respective lanes and / or lanes and by relating the respective lanes and / or lanes to one another in terms of their arrangement between the different traffic participants, possible interactions between the traffic participants can be taken into account.

[0044] By taking into account the information that two road users are located successively on the same lane and / or lane or side by side on different lanes and / or lanes or on two intersecting lanes and / or lanes, the respective behavior of the respective road users can be taken into account.

[0045] Depending on the arrangement of the traffic participants on the same lane and / or lane, on lanes and / or lanes running side by side or on intersecting lanes and / or lanes, different behaviors of the traffic participants which are however characteristic for the respective situation can be expected.

[0046] For example, in the case of traffic participants arranged one behind the other in a lane and / or lane, it can be predicted that the following and faster traffic participant in the direction of travel will reduce speed in the future or, if possible, change lane and / or lane.

[0047] In contrast, in the case of traffic participants who are situated on lanes and / or lanes arranged next to each other, a reduction in speed or a change of lanes and / or lanes cannot necessarily be expected.

[0048] In contrast, in the case of traffic participants on crossing lanes and / or roadways, a reduction in speed when entering the intersection region of lanes and / or roadways is a very typical and therefore possible behavior.

[0049] Therefore, by considering the corresponding relative arrangement relationship of the interactive graphical representation, further refinement of the trajectory prediction can be achieved.

[0050] According to one embodiment, the prediction of the motion trajectory includes:

[0051] determining, for a traffic participant in the surroundings of the own vehicle, a distance between the traffic participant and other traffic participants and / or the own vehicle based on the interactively graphically represented position information and lane information;

[0052] Based on the distance and speed information of the traffic participants, determining the probability value of the traffic participants colliding with each other or with the vehicle;

[0053] In this case, the movement trajectory is completed while taking into account the determined probability value.

[0054] The following technical advantage can be achieved in this way: By determining the distance and the probability value of a collision between the road users based thereon, the trajectory prediction can be further improved. It is assumed that the road users will perform maneuvers with a certain probability in order to prevent a collision of this type. By taking into account possible maneuvers of this type, a more precise prediction of the future movement trajectory of the respective road users can be achieved.

[0055] According to one specific embodiment, the nodes of the interactive graphical representation include information about the type of the respective road user, wherein the type is defined as one of the following list: vehicle, bus, truck, motorcyclist, cyclist, pedestrian, animal.

[0056] This can achieve the following technical advantage: To determine the interaction between traffic participants, the corresponding type can be additionally taken into account. This allows for a further refinement of the trajectory prediction. Different types of traffic participants generally have different characteristic behaviors. By taking these different characteristic behaviors into account, the trajectory prediction can be further improved.

[0057] According to one embodiment, the prediction module is constructed as an artificial intelligence that is trained to implement the above-mentioned method steps, wherein the prediction module includes a participant level, a map level, a fusion level, a merger level and at least one encoder element, wherein the participant level is configured to process trajectory data and / or generate an interactive graphical representation, wherein the map level is configured to process map data, wherein the fusion level is configured to merge results of the participant level and the map level, wherein at least one encoder element is configured to process further map data configured as image data and to extract lane features and / or lane features from the map data, wherein the merger level is configured to predict and output at least one movement trajectory of at least one traffic participant based on results of the fusion layer and results of the encoder element.

[0058] The following technical advantages can be achieved: a prediction module with excellent performance can be provided. In particular, the prediction module can be trained to implement the above method steps. This allows high accuracy of trajectory prediction.

[0059] According to one specific embodiment, the artificial intelligence comprises at least one graphical neural network and / or the encoder element is designed as an autoencoder and is configured to generate a latent space representation in a latent space of roadway features based on image data of the further map data.

[0060] The following technical advantages can be achieved: a prediction module with excellent performance can be provided. By configuring the encoder element as an autoencoder, it can be caused that the image information of the image data of the further map data is accurately and reliably encoded into the latent space representation. This enables accurate processing of the image information of the further map data.

[0061] According to one aspect, there is provided a method for controlling a host vehicle, the method comprising:

[0062] Implementing trajectory prediction according to the method for trajectory prediction according to one of the above embodiments;

[0063] At least one control function of the host vehicle is executed taking into account the predicted driving trajectory of the vehicle.

[0064] As a result, the following technical advantage can be achieved: an improved method for controlling a host vehicle can be provided, which is based on the improved method for trajectory prediction having the above-mentioned technical advantages.

[0065] According to one aspect, a computing unit is provided which is configured to carry out a method for trajectory prediction and / or a method for controlling a host vehicle according to one of the above-described specific embodiments.

[0066] According to one aspect, a computer program product is provided, which includes instructions. When the program is executed by a data processing unit, the instructions cause the data processing unit to implement the method for trajectory prediction and / or the method for controlling a vehicle according to one of the above-mentioned embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0067] Embodiments of the present invention are described with reference to the following drawings. The drawings show:

[0068] Figure 1 A schematic diagram showing a system for trajectory prediction;

[0069] Figure 2 A schematic diagram showing an interactive graphical representation according to one embodiment;

[0070] Figure 3 A graphical illustration showing the generation of an anchor path according to one embodiment;

[0071] Figure 4A schematic diagram showing a prediction module according to one embodiment;

[0072] Figure 5 A flow chart showing a method for trajectory prediction;

[0073] Figure 6 A flow chart showing a method for controlling a host vehicle;

[0074] Figure 7 A schematic diagram illustrating a computer program product. DETAILED DESCRIPTION

[0075] Figure 1 A schematic diagram of a system 200 for trajectory prediction is shown.

[0076] In the illustrated embodiment, the system 200 for trajectory prediction includes at least one host vehicle 201, which has a computing unit 255. A prediction module 209 and an environment detection device 257 are installed on the computing unit 255. In addition, the host vehicle 201 includes at least one environment sensor 237. The environment sensor 237 can record environment sensor data 235 of the environment of the host vehicle 201, based on which the environment detection device 257 can perform detection of the environment.

[0077] Furthermore, prediction module 209 is configured to carry out the method according to the present invention for trajectory prediction.

[0078] In the illustrated diagram, an exemplary traffic scenario is shown in order to illustrate the method for trajectory prediction. In the exemplary traffic scenario, an intersection area of ​​two intersecting lanes 223 is shown. The two lanes 223 each have two lanes 224 arranged side by side. In the environment of the vehicle 201, four other traffic participants 207 are arranged in this traffic scenario. The traffic participants 207 are arranged on different lanes 223 or lanes 225.

[0079] In order to carry out trajectory planning of the movement trajectory 205 of the further road users 207, the prediction module 209 first receives trajectory data 203 of at least one movement trajectory 205 of at least one of the further road users 207. According to the present invention, the trajectory data 203 are formed in a graphical representation 211. The graphical representation 211 includes a plurality of nodes 213 and a plurality of edges 215 connecting the nodes 213 for illustrating the movement trajectory 205.

[0080] Here, the node 213 includes the position information and / or speed information of the corresponding traffic participant 207 at a given point in time. In contrast, the edge describes the time relationship of the nodes 213 connected by the corresponding edge 215. Therefore, the edge 215 connecting two nodes 213 defines the following time sequence: the position information and / or speed information of the corresponding nodes 213 is recorded in this time sequence.

[0081] Furthermore, map data 217 representing a map representation 219 of the environment of host vehicle 201 are received by prediction model 209. According to the invention, map data 217 are written in a graphical representation 211. Graph representation 211 comprises a plurality of nodes 213 and edges 215 connecting nodes 213.

[0082] Here, the map representation 219 is an electronic road map of the environment of the host vehicle. Here, the electronic road map can be constructed in a format similar to the NuScenes data set. In addition, the road map can also be provided with lane segments and / or lane segments according to the Lanelet2 format.

[0083] Node 213 of map data 217 includes position information of lane boundary elements 221 of lane 223 and / or lane 225 on which respective road user 207 is traveling.

[0084] The edge 215 includes a relative arrangement relationship related to position between the nodes 213 connected via the edge. Therefore, the edge 215 can define how the individual position information of the nodes 213 are arranged relative to each other. Here, the position information defines the individual parts of the corresponding lane boundary element 221. Here, the edge 215 defines how the position information is arranged relative to each other, for example, arranged one after another or side by side.

[0085] For a detailed description of the construction of the trajectory data 203 and the map data 217 in the corresponding graphical representation 211, reference is made to the literature of Grimm et al.: D. Grimm, P. M. Dreβler, J.-M. "Holistic graphbased motion prediction", 2023IEEE International Conference on Robotics an Automation (CRA), 2023, pp. 2965-2972. In the literature of Grimm et al., nodes 213 of trajectory data 203 are called "agent-nodes". In the literature of Grimm et al., nodes 213 of map data 217 are called "map-nodes".

[0086] In the specific embodiment shown, lane boundary element 221 is given by the center line of lane 223 or lane 225 , in which the respective road user 207 for whom future movement trajectory 205 is to be predicted is driven.

[0087] Alternatively, further lane boundary elements 221 , such as edge boundaries or lane markings, may be taken into account.

[0088] According to the present invention, an interaction graphic representation 227 is generated based on trajectory data 203 and map data 217. Interaction graphic representation 227 describes interactions between traffic participants 207. Interactions are shown in interaction graphic representation 227 by edges 215. Nodes 213 of interaction graphic representation 227 show traffic participants 207 arranged in the environment of host vehicle 210.

[0089] The edges 215 connecting the nodes 214 in the interactive graphical representation 227 define the position-dependent relative arrangement of the road users 207 represented by the respective nodes 213 relative to one another. In this case, the edges 215 connecting the nodes 214 define whether the road users 207 represented by the nodes 213 are: arranged successively on the same lane 223 or lane 225; on different lanes 223 or lanes 225 arranged side by side or on crossing lanes 223 or lanes 225.

[0090] For a detailed description of the interactive graphical representation 227, refer to Figure 2 Description.

[0091] According to the present invention, based on trajectory data 203 , map data 217 and interactive graphical representation 227 , prediction module 209 predicts a movement trajectory 205 to be executed in the future by respective road user 207 for at least one road user 207 in the environment of host vehicle 201 .

[0092] In this case, movement trajectory 205 is predicted in such a way that predicted movement trajectory 205 matches trajectory data 203 of the corresponding road user 207. This is to be understood as meaning that trajectory data 203 of movement trajectory 205 already performed by road user 207 and predicted movement trajectory 205 of the same road user 207 form a continuously extending overall trajectory.

[0093] Furthermore, movement trajectory 205 is predicted taking into account map data 217 such that the predicted movement trajectory follows the course of lane 223 or lane 225 defined by map data 217 . This may include lane changes and / or lane changes by the respective road user 207 .

[0094] For a detailed description of the use of trajectory data 203 and map data 217 for trajectory prediction, reference is again made to Grimm et al.

[0095] By taking into account the information of the interactive graphical representation, the interactions of different road users 207 may also be taken into account for the trajectory prediction.

[0096] This is based on the idea that, depending on the respective road user 207 on the same lane 223 or lane 225 , on lanes 223 or lanes 225 arranged next to each other, or on intersecting lanes 223 or lanes 225 , road users 207 display different behaviors that are characteristic of the respective arrangement.

[0097] This arrangement-related behavior may be learned by the prediction module 208 and thus simultaneously incorporated into the trajectory prediction.

[0098] Furthermore, further map data 253 may be taken into account by prediction module 208 . In this case, further map data 253 are in the form of image data and include information about roadway characteristics 239 .

[0099] Lane characteristics 239 may include, for example, the presence of traffic signs, traffic lights, crosswalks, bus stops, construction sites, or similar permanent characteristics of lane 223 or lane 225 in the environment of the vehicle, which may influence the behavior of traffic participants 207 in the environment of the vehicle 201 .

[0100] Prediction module 209 is configured to extract lane features 239 from further map data 253 and to take lane features 239 into account in the trajectory prediction.

[0101] According to one specific embodiment, prediction module 209 is further configured to generate an anchor path for trajectory prediction. The anchor path describes the spatial region in which possible movement trajectories 205 of road users 207 must be arranged. The anchor path thus forms a limit for possible movement trajectories 205 of road users 207. Prediction module 209 is configured to exclude movement trajectories 205 that extend at least partially outside the anchor path as invalid movement trajectories 205 in the trajectory prediction.

[0102] For a detailed description of anchor paths, refer to Figure 3 Description.

[0103] Furthermore, prediction module 209 may be configured to determine a probability value for a collision between road users 207 or between road users 207 and host vehicle 201 based on speed information of road users 207 and based on distance information between road users 207 or between road users 207 and host vehicle 201 .

[0104] In this case, speed information of road user 207 can be obtained from trajectory data 203 . Distance information can be obtained from surroundings detection device 257 based on surroundings sensor data 234 .

[0105] Furthermore, the relative arrangement of road users 207 in interactive graphical representation 227 may be taken into account in order to calculate the probability value for a collision.

[0106] The respectively determined probability value of a collision of road user 207 can be incorporated into the trajectory prediction by assuming, in the case of the traffic behavior of road user 207 to be predicted, that road user 207 behaves in such a way that a collision is avoided.

[0107] According to one specific embodiment, prediction module 209 is also configured to simultaneously take into account the classification of road user 207 in the trajectory prediction.

[0108] In this classification, road user 207 may be identified, for example, as a vehicle, a motorcyclist, a cyclist, a pedestrian, or an animal.

[0109] This type of classification can be used, for example, to generate an anchor path. According to one specific embodiment, an anchor path is calculated only for those road users 207 that can be driven mainly or only on lane 223 or lane 225. Thus, an anchor path is not calculated for pedestrians and animals. If necessary, the determination of the anchor path can be limited to vehicles only, so that an anchor path is not generated even for motorcyclists and / or cyclists.

[0110] Alternatively, the classification of vehicles can be further refined and a distinction can be made between passenger cars, trucks, buses or trams. This distinction allows a further differentiation of the possible behaviors of the vehicles. This is based on the following reasonable assumption: trucks can usually be expected to behave differently from passenger cars, for example because trucks cannot travel on bridges or underpasses that are easy for passenger cars to travel on, or because different traffic regulations apply to trucks than to passenger cars, such as speed limits or overtaking regulations.

[0111] In addition, further differentiation between different types of passenger vehicles may also be considered in order to be able to further differentiate the driving behavior, for example as active or passive, sporty or conservative.

[0112] Furthermore, a classification of road users 207 and the expected traffic behavior may be taken into account in the trajectory prediction in order to thereby achieve a further refinement of the expected movement trajectory.

[0113] The classification of road user 207 may be realized by surroundings detection device 257 based on surroundings sensor data 235 of surroundings sensor 237. In this case, surroundings detection device 257 may be designed as a trained artificial intelligence.

[0114] Figure 2 A schematic diagram of an interactive graphical representation 227 is shown according to one embodiment.

[0115] In Figure a) it is shown again Figure 1 2, wherein an intersection of a first lane 279 and a second lane 281 is shown. Here, the first lane 279 includes a first lane 283 and a second lane 285. The second lane 281 includes a third lane 287. The first and second lanes 279, 281 intersect in the intersection area. Here, in particular, the second lane 285 of the first lane 279 intersects with the third lane 287 of the second lane 281.

[0116] In contrast, first lane 283 of first lane 279 has a direction change region 289 and, after the direction change, transitions into third lane 287 of second lane 281 .

[0117] The first road user 259 is arranged in the first lane 283 of the first lane 279. The second road user 261 is arranged in the third lane 287 of the second lane 281. The third road user 263 is also arranged in the third lane 287 of the second lane 281, but in the driving direction behind the second road user 261 and behind the intersection area of ​​the two lanes 279, 281. The fourth road user 265 is arranged in the second lane 285 of the first lane 279 and in the driving direction in front of the intersection area.

[0118] Graph b) shows an interactive graphic arrangement 227 depicting the traffic scene of graph a). In the interactive graphic representation 227, traffic participants 259, 262, 263, 265 and the host vehicle 201 are shown by nodes 213. The relationships between the traffic participants 259, 262, 263, 265 and the relationship between the traffic participants and the host vehicle 201 are defined by edges 215, via which at least two nodes 213 are respectively connected to one another. Three different arrangements 267, 269, 271 are defined by the edges 215.

[0119] Arrangement 267 relates to the situation in which at least two road users 207 and / or own vehicle 201 are arranged one after the other on the same lane 223 and / or lane 225 .

[0120] Arrangement 269 describes a situation in which at least two road users 207 and / or own vehicle 201 are arranged side by side on different lanes 223 and / or lanes 225 .

[0121] Arrangement 271 describes a situation in which at least two road users 207 and / or own vehicle 201 are arranged on intersecting lanes 223 and / or lanes 225 .

[0122] In the traffic scene of figure a), the first road user 259 is arranged in front of the direction change region 289 on the first lane 283 of the first lane 279. The second road user 261 is arranged behind the direction change region 289 on the third lane 287 of the second lane 281. In this area, the first lane 283 of the first lane 279 and the third lane 287 of the second lane 281 transition into each other. Therefore, the first and second road users 259, 261 are arranged in the same lane.

[0123] In diagram b), the first and second road users 259, 261 are therefore connected to one another only via arrangement 267, which defines a successive arrangement on the same roadway. The direction of the arrows of arrangement 267 indicates the order of the arrangement of the two road users 259, 261 relative to the direction of travel. As shown in diagram a), the second road user 261 is arranged in front of the first road user 259 in the direction of travel.

[0124] Similarly, the second road user 261 and the third road user 263 are arranged one after the other on the third lane 287 of the second lane 281 . In the diagram b), the two road users 261 , 263 are respectively connected to one another only via the arrangement 267 .

[0125] In contrast, the third road user 263 and the fourth road user 265 are arranged on the third lane 287 of the second lane 281 and the second lane 285 of the first lane 279, which intersect each other. In the diagram b), the two road users 261, 263 are respectively connected to each other via an arrangement 271, which defines an arrangement on intersecting lanes or lanes. In the case of an arrangement on intersecting lanes or lanes, the order of the road users 207 with respect to the direction of travel is not taken into account. Therefore, the third and fourth road users 263, 265 are connected to each other by two opposing arrows of the arrangement 271.

[0126] Accordingly, first road user 259 and third road user 263 are arranged on first lane 283 of intersecting first lane 279 and third lane 287 of second lane 281. Since third road user 263 is located in third lane 287 of second lane 281 in front of the intersection area and thus in front of the area where first lane 283 of first lane 279 transitions into third lane 287 of second lane 281, first road user 259 and third road user 263 are arranged on intersecting lanes 223 or lanes 225.

[0127] Correspondingly, in the graphical representation in diagram b), road users 259 , 263 are connected by an arrangement 271 in the form of two arrows pointing in opposite directions.

[0128] According to one embodiment, the position determination of the host vehicle 201 is performed in a manner with a probability value attached. Therefore, in order to determine the position of the host vehicle 201, a plurality of results with a probability value attached can be obtained. This is shown in the diagram a) by two dashed markings. The dashed markings define a first position 291 on the first lane 283 of the first lane 279 and a second position 293 on the second lane 285 of the first lane 279 of the host vehicle 201.

[0129] In the traffic scenario shown, host vehicle 201 is arranged to a greater extent in second lane 285 of first lane 279. Therefore, second position 293 is to be provided with a greater probability value than first position 291.

[0130] Alternatively or additionally, more than two positions with corresponding probability values ​​may also be output as the result of the position determination.

[0131] Depending on the probability values ​​of the first and second positions 293, 293, the host vehicle 201 is arranged on the same first lane 283 as the first road user 259 or on the second lane 285 and thus next to the first road user 259. Accordingly, in the interactive graphical representation 227 of the diagram b), the two nodes 213 of the host vehicle 201 and the first road user 259 are connected to one another via two arrangements 267, 269.

[0132] Accordingly, host vehicle 201 is arranged in first position 291 on the same lane 283 as second road user 261, and in second position 293 on second lane 285 and thus next to second road user 261. Thus, in interactive graphical representation 227, nodes 213 of host vehicle 201 and of second road user 261 are connected via arrangements 267, 269.

[0133] The same applies to the fourth road user 265. Due to the two possible positions 291, 293, the host vehicle 201 is either arranged on the same lane 285 as the fourth road user 265 or on a lane 283 arranged next to it. In the interactive graphic representation 227 of the diagram b), the nodes 213 of the host vehicle 201 and the fourth road user 265 are connected via the arrangements 267, 269.

[0134] With regard to the third road user 263, there is no distinction between the two positions 291, 293 of the host vehicle 201. In both positions 291, 293, the host vehicle 201 is arranged on lanes 283, 285 intersecting the third lane 287 of the second lane 281. Therefore, in the interactive graphical representation 227 of the diagram b), the nodes 213 of the host vehicle 201 and the third road user 263 are connected only via the arrangement 271.

[0135] Furthermore, the diagram a) shows a distance 241 between own vehicle 201 and road users 207 or between road users 207. Figure 1 As described, based on distance 241 and speed values ​​of own vehicle 201 and / or road users 207 , a probability value of a possible collision of road users 207 with one another or with own vehicle 201 may be calculated taking into account the arrangement of interactive graphical representation 227 .

[0136] In diagram a), distance 241 between own vehicle 201 and / or road users 207 is defined by the center point of one road user 207 to the center point of the respective other road user 207 of own vehicle 201. However, other distance definitions are also conceivable.

[0137] The traffic scenarios shown in the diagrams a) and b) are merely exemplary. According to the present invention, any desired traffic scenarios can be taken into account in the interactive graphic representation 227 .

[0138] Based on the received trajectory data 203 and map data 217, the position of the traffic participant 207 on the corresponding lane 223 and / or lane 225 in the environment of the host vehicle 201 is determined. Through the position information of the trajectory data 203 and the information represented by the map, the traffic participant 207 can be recognized as being located on the corresponding lane 223 or lane 225.

[0139] As described above, trajectory data 203 and in particular position information of trajectory data 203 may be generated by analyzing surrounding sensor data 235 and surrounding detection device 257. Surrounding detection device 257 may detect and locate road users 207 in the surroundings and on lane 223 or lane 225.

[0140] Alternatively or additionally, trajectory data 203 of road user 207 may be transmitted directly from road user 207 to host vehicle 201 via car-2-car communication.

[0141] By comparing with the information represented by the map, road user 207 can be recognized as being located on the corresponding lane 223 or lane 225 .

[0142] The information of interactive graphical representation 227 may be incorporated into the trajectory prediction as follows: In this case, it is assumed that the behavior of different road users 207 situated in the environment of own vehicle 201 is related to the relative behavior of road users 207 with respect to one another.

[0143] For example, when two road users 207 are arranged in the same lane 225 , if rear road user 207 is faster than front road user 207 , a reduction in the speed of rear road user 207 or a lane change by rear road user 207 can be assumed.

[0144] In contrast, if two road users 207 are situated in two lanes 225 arranged next to each other, a speed change or a lane change by one of road users 207 cannot be directly anticipated.

[0145] In contrast, if two road users 207 are situated on intersecting lanes 223 or lanes 225 , a speed reduction may be assumed before entering the intersection area.

[0146] Therefore, through the additional information of the interactive graphical representation 227 , trajectory prediction may be improved.

[0147] For a detailed description of the generation and use of interactive graphical representations for trajectory prediction, see Zipfl et al., M. Zipfl, F. Hertlein, A. Rettinger, S. Thoma, L. Halilaj, J. Luettin, S. Schmid, C. Henson, “Relation-based Motion Prediction using Traffic Scene Graphs,” arXiv:2212.02503. In Zipfl et al., the interactive graphical representation is referred to as a “traffic scene graph.”

[0148] Figure 3 A graphical illustration of the generation of anchor path 229 is shown.

[0149] Figure 3 The generation of anchor path 229 in a method for trajectory prediction is shown. Anchor path 229 is generated only for road users 207 that are identified as vehicles of any design and are recognized as being located on lane 223 or lane 225 .

[0150] In order to generate the anchor path 229, a map representation 219 of a road traffic network constructed in the environment of the host vehicle 201 is first provided. In this case, the map representation 219 is constructed as an electronic road map, depicting lanes 223 and / or lanes 225 in the environment of the host vehicle 201, and showing lane segments 231 of the lanes 223 and / or lane segments 233 of the lanes 225 in the Lanelet2 format.

[0151] Lane segments 231 or lane segments 233 in the Lanelet2 format each include a node 295 and at least two line elements 297. Line element 297 includes node 295 and shows the right boundary or left boundary of the corresponding lane segment 231 or lane segment 233, respectively. Lane segment 231 or lane segment 233 depicts lane 223 or lane 225 in map representation 219, respectively. Lane segment 231 or lane segment 233 can depict the entire lane 223 or lane 225 or only a part of lane 223 or lane 225. Line element 297 depicts the edge boundary of the depicted lane 223 or lane 225.

[0152] Furthermore, lane segment 231 or lane segment 233 may include a lane identification element or a lane identification element. The lane identification element or the lane identification element may be used to clearly and uniquely identify the corresponding lane segment 231 or lane segment 233.

[0153] At least one road user 207 is then ascertained in the surroundings of host vehicle 201. Host vehicle 201 is not shown in diagrams a) and b).

[0154] The respective road user 207 is then mapped onto a lane segment 231 or lane segment 233 of map representation 219 , in which road user 207 is situated at the time at which the generation of anchor path 229 is carried out.

[0155] A lane anchor point or lane anchor point is then determined. The lane anchor point or lane anchor point includes a plurality of further lane segments 231 or lane segments 233 which are connected in map representation 219 to lane segment 231 or lane segment 233 in which road user 207 is arranged, and in which movement trajectory 205 may extend.

[0156] In this case, movement trajectory 205 may extend through the respective lane segment 231 or lane segment 233 if this results in a movement trajectory 205 having a coherent or continuous course that complies with the applicable traffic regulations and is suitable for controlling the respective road user 207 .

[0157] For this purpose, a lane anchor point or a lane anchor point may respectively include a plurality of lane marking elements or lane marking elements.

[0158] The lane anchor points or lane anchor points may be limited to a maximum length, for example 100 meters.

[0159] At least one possible movement trajectory 205 of road user 207 is then ascertained, which extends at least over a portion of a further roadway segment 231 or lane segment 233 that is connected to roadway segment 231 or lane segment 233 in which road user 207 is situated.

[0160] To this end, the lane anchor or lane anchor can be used to generate the anchor path 229 and the 128-dimensional anchor embedding. Here, the anchor embedding is the latent space representation (Latent-Raum- ) and is generated by a prediction module 209 which is constructed as an artificial intelligence, in particular a graph neural network GNN. In this case, the latent space representation is a vector representation in the latent space of the artificial intelligence. In this case, the anchor embedding can be generated by the GNN using the position information of the trajectory data 203, the map data 217 and the lane anchor points or lane anchor points.

[0161] In the trajectory decoder of the GNN, the information of the traffic participant 207 may be copied k times, where k∈[0,9]. Then, the information of the traffic participant 207 may be linked with the anchor embedding and used for trajectory prediction.

[0162] T=MPLP(concat(z Teilnehmer , z Anker )

[0163] Here, T is the motion trajectory 205, z Teilnehmer is the latent space representation of the corresponding traffic participant 207, and Z Anker is the latent space representation or anchor embedding of the lane anchor or lane anchor.

[0164] exist Figure 3 2 shows a plurality of possible movement trajectories 205 that can satisfy anchor path 229 and can be implemented by road user 207. In addition, a plurality of possible movement trajectories 205 can be reduced to a predetermined number of as diverse as possible movement trajectories 205. In this case, redundant movement trajectories 205 can remain out of consideration.

[0165] Anchor paths 229 generated in this way serve as spatial boundaries for possible movement trajectories 205 . In the trajectory prediction, movement trajectories 205 predicted for different road users 207 must lie within anchor paths 229 generated for the respective road users 207 .

[0166] By taking anchor path 229 into account, prediction module 209 is trained to only generate movement trajectories 205 that extend within anchor path 229 generated for the respective road user 207 .

[0167] The anchor path 229 can be constructed as a polyline with 2D position information.

[0168] Figure 4 A schematic diagram of the prediction module 209 according to one embodiment is shown.

[0169] According to one specific embodiment, prediction module 209 is designed as an artificial intelligence that is trained to carry out the method for trajectory prediction according to the present invention.

[0170] For this purpose, the prediction module 209 may be configured, for example, as a correspondingly trained graphical neural network GNN.

[0171] In the illustrated embodiment, prediction module 209 includes a participant level 243 , a map level 245 , a fusion level 247 , a merge level 249 , and two encoder elements 251 .

[0172] Trajectory data 203 are processed in a graphical representation via participant level 243 . Furthermore, in participant level 243 , an interactive graphical representation 227 is generated based on trajectory data 203 and taking into account cartographic information about the course and position of the respective lanes 223 and lanes 225 .

[0173] In the embodiment shown, participant level 243 includes a first sub-level 273 and a second sub-level 275. In first sub-level 273, trajectory data 203 of different traffic participants 207 are processed. In second sub-level 275, based on the results of first sub-level 273, interactive graphical representations 227 are generated.

[0174] In the map layer 245 , the map data 217 is processed in a graphical representation.

[0175] In the fusion level 247, the results of the participant level 243 and the map level 245 are merged. In the fusion level 247, an anchor path 229 is also generated based on the results of the participant level 243 and the map level 245.

[0176] Encoder element 251 is configured to read in further map data 251 in an image format and to extract lane features 239. Encoder element 251 may be designed, for example, as part of an autoencoder having encoder element 251 and a corresponding decoder element.

[0177] In the merging level 249, the results of the fusion level 247 and the encoder elements 251 are processed and based on this, the movement trajectory 205 is predicted. Thus, the trajectory prediction is performed based on the trajectory data 203, the map data 217, and taking into account the information of the interactive graphic representation 227, the anchor path 229 and the lane characteristics 239. The predicted movement trajectory 205 is output as output data 277 via the merging level 249.

[0178] The image data of further map data 253 may be based on a map representation similar to the NuScenes data set. Here, by means of the autoencoder of encoder element 251 , lane segments 231 and / or lane segments 233 may also be integrated into the map representation in the sense of the Lanelet2 format.

[0179] Furthermore, a 10-channel image of the image data of further map data 253 can be converted into a latent space representation by an autoencoder of encoder element 251 . In this case, a latent space representation is understood to be a vector representation of the information of further map data 253 in the latent space of encoder element 251 .

[0180] In this case, the architecture of the autoencoder of encoder element 251 may be designed such that encoder element 251 is set up to encode 10 binary masks of the image data of further map data 253 in the format 10×128×128 into a 128-dimensional latent space representation.

[0181] To this end, the encoder element 251 may include six layers, wherein each layer includes a 2D convolution, batch normalization, and a LeakyReLu activation.

[0182] Correspondingly, the decoder element of the autoencoder can also include six layers, wherein the first five layers include 2D expansion (Entfaltungen), batch normalization and ReLu activation respectively. In the last layer, the Tanh function can be used as the activation function.

[0183] The 128×128-dimensional image data of the additional map data 253 can cover a spatial area of ​​50 m×50 m of the environment of the host vehicle 201 .

[0184] Can be 2.10 -4 The autoencoder is trained for 100 epochs with a learning rate of 1 and an MSE reconstruction loss of 0.59.

[0185] Furthermore, nodes 213 of the graphical representation of trajectory data 203 may be provided with additional map information. For this purpose, a latent spatial representation of a local segment area of ​​the map representation of the environment of the respective road user 207 may be added to the node 213 representing the respective road user 207 in the graphical representation of trajectory data 203.

[0186] n * =MLP(concat(n, dropout(z map )))

[0187] In this case, n* is the node 213 of the respective road user 207 for which map information is provided, while n describes the original node 213. In contrast, zmap describes the segment area of ​​the map representation.

[0188] To avoid overfitting, a packet loss rate (drop-out rate) of 0.5 may be used. In this case, the segment area may be centered around the corresponding road user 207 in the map representation and rotated in such a way that the direction of movement points to the north.

[0189] Furthermore, the trajectory prediction by prediction module 209 designed as a trained artificial intelligence is carried out with the aid of a loss function. In this case, the loss function can be defined as follows:

[0190] L=L reg +w1L score +w2L yaw

[0191] Here, L is the loss function, L reg is the regression loss, L score is the score loss, L yaw is the orientation loss, and w1, w2 are parameters. Here, the regression loss can be constructed as a continuous L1 loss. The score loss can be constructed as a maximum margin loss (Max-Margin-Loss). Here, the regression loss and the score loss are evaluated only for the predicted motion trajectory 205 with the smallest final displacement error relative to the ground truth trajectory.

[0192] By comparing the orientation of the predicted motion trajectory 205 with the orientation of the anchor path 229 , the prediction module 209 can be made to predict only the motion trajectory 205 that follows the anchor path 229 .

[0193] The orientation loss can also be expressed as follows:

[0194] L yaw =1-cos(θ Anker -θ)

[0195] Here, the angle θ is defined within (-π, π].

[0196] Prediction module 209 Figure 4 The architecture shown in is only exemplary. The architecture can also be configured in different ways.

[0197] Figure 5 A flow chart of a method 100 for trajectory prediction is shown.

[0198] In a first method step 101, trajectory data 203 of a movement trajectory 205 of a road participant 207 arranged in the environment of the host vehicle 201 are first received by a prediction module 209. In this case, the trajectory data 203 are arranged in a graphical representation 211, wherein nodes 213 of the graphical representation 211 include position information and / or speed information of the road participant 207, wherein edges 215 of the graphical representation 211 define time relationships between the position information and / or speed information of the road participant 207.

[0199] In a further method step 103, map data 217 of a map representation 219 depicting the environment of host vehicle 201 are received by prediction model 209. In this case, map data 217 are arranged in a graphical representation 211, wherein nodes 213 of graphical representation 211 include position information of lane boundary elements 221 of lanes 223 and / or lanes 225 traveled by host vehicle 201 and / or by other road users 207. In this case, edges 215 of graphical representation 211 define spatial relationships between the position information of lane boundary elements 221.

[0200] In a further method step 105, an interactive graphic representation 227 is generated for a plurality of traffic participants 207 by means of a prediction model 209 based on the trajectory data 203 of the traffic participants 207 and the lane pose information of the map representation 219. In this case, the nodes 213 of the interactive graphic representation 227 each show a traffic participant 207 located in the environment of the host vehicle 201 and include position information and / or speed information of the traffic participant 207. The edges 215 of the interactive graphic representation 227 each define an arrangement relationship between the traffic participants 207 shown by the nodes 213.

[0201] In a further method step 117 , further road users 207 are classified by an environment detection device into one of the following list: vehicle, bus, truck, tram, motorcyclist, cyclist, pedestrian, animal based on environment sensor data 235 of at least one environment sensor 237 of host vehicle 201 .

[0202] In a further method step 119, further map data 253 are received by prediction model 209, which depict a map representation 219 of the surroundings of host vehicle 201. In this case, further map data 253 are in the form of image data.

[0203] In a further method step 121 , lane characteristics 239 of lane 223 on which road user 207 is travelling are read from further map data 253 by predictive model 209 .

[0204] In a further method step 109, an anchor path 229 is generated by prediction model 209 based on trajectory data 203 of road user 207 and map data 217 of map representation 219. Anchor path 229 defines an area on lane 223 traveled by road user 207 in which possible movement trajectories 205 can be arranged.

[0205] For this purpose, in a further method step 111 , at least one road user 207 is identified by prediction module 209 on lane 223 or lane 225 .

[0206] In a further method step 113, lane segments 231 and / or lane segments 233 of lanes 223 and / or lanes are identified by prediction module 209 on which road user 207 is located. Lane segments 231 and / or lane segments 233 define sub-areas of lanes 223 and / or lanes 225 and are bounded by lane boundary elements 221 of lanes 223 and / or lane boundary elements of lanes 225. In addition, lane segments 231 can be connected to further lane segments 231 and / or lane segments 233 can be connected to further lane segments 233. In this case, a plurality of lane segments 231 and / or a plurality of lane segments 233 depict lanes 223 and / or lanes 225.

[0207] In a further method step 115 , lane segments 231 and / or lane segments 233 are connected to further lane segments 231 and / or lane segments 233 that are respectively connectable to one another to form anchoring path 229 , taking into account the prevailing traffic regulations and the course of lane 223 and / or lane 225 .

[0208] In a further method step 107 , a prediction is made by means of a prediction model 209 based on the trajectory data 203 , the map data 217 , an interactive graphical representation 227 of the road user 207 and further map data.

[0209] For this purpose, in a further method step 123 , a distance 241 of road user 207 from other road users 207 and / or from ego vehicle 201 is ascertained for road user 207 in the surroundings of ego vehicle 201 based on the position information and the lane information of interactive graphical representation 227 .

[0210] In a further method step 125 , a probability value of a collision of the road users with one another or with own vehicle 201 is calculated based on distance 241 and speed information of road users 207 . In this case, the probability value is taken into account when predicting movement trajectory 205 .

[0211] Figure 6A flow chart of method 100 for controlling host vehicle 201 is shown.

[0212] In a first method step 301 , firstly a trajectory prediction is carried out according to method 100 for trajectory prediction for a vehicle according to one of the above-described specific embodiments.

[0213] In a further method step 303 , at least one control function of host vehicle 201 is then carried out taking into account predicted movement trajectory 205 of host vehicle 201 .

[0214] Figure 7 A schematic diagram of a computer program product 400 is shown, which includes instructions which, when the program is executed by a data processing unit, cause the data processing unit to carry out method 100 for trajectory prediction and / or method 300 for controlling host vehicle 201 .

[0215] In the embodiment shown, computer program product 400 is stored on a computing unit or storage medium 401. Storage medium 401 may be any desired storage medium known from the prior art.

Claims

1. Computer-implemented method for trajectory prediction (100): Trajectory data (203) of a movement trajectory (205) of a traffic participant (207) arranged in the environment of the host vehicle (201) are received (101) by a prediction module (209), wherein: The trajectory data (203) are arranged in a graphical representation (211), wherein nodes (213) of the graphical representation (211) include position information and / or speed information of the traffic participants (207), wherein edges (215) of the graphical representation (211) define time relationships between the position information and / or speed information of the traffic participants (207); receiving (103) map data (217) of a map representation (219) depicting an environment of the host vehicle (201) by means of the prediction module (209), wherein the map data (217) are arranged in a graphical representation (211), wherein nodes (213) of the graphical representation (211) include position information of lane boundary elements (221) of lanes (223) and / or lanes (225) traveled by the host vehicle (201) and / or by further traffic participants (207), wherein edges (215) of the graphical representation (211) define spatial relationships between the position information of the lane boundary elements (221); An interactive graphical representation (227) is generated (105) for a plurality of traffic participants (207) based on the trajectory data (203) of the traffic participants (207) and the lane posture information of the map representation (219) by means of the prediction module (209), wherein each node (213) of the interactive graphical representation (227) shows a traffic participant (207) located in the environment of the host vehicle (201) and includes position information and / or speed information of the traffic participant (207), and wherein each edge (215) of the interactive graphical representation (227) defines an arrangement relationship between two of the traffic participants (207) shown by the node (213); The prediction module (209) predicts (107) a future movement trajectory to be carried out for at least one further road user (207) based on the trajectory data (203), the map data (217) and the interactive graphical representation (227) of the road user (207).

2. The method (100) according to claim 1, further comprising: An anchor path (229) is generated (109) by the prediction module (209) based on the trajectory data (203) of the traffic participant (207) and the map data (217) of the map representation (219), wherein the anchor path (229) defines the following area on the lane traveled by the traffic participant (207): possible movement trajectories (205) can be arranged in the area; wherein the movement data is predicted based on the trajectory data (203), the map data (217), the interactive graphic representation (227) of the traffic participant (207) and / or the anchor path (229).

3. The method (100) according to claim 2, wherein: The generation (109) of the anchor path (229) includes: At least one traffic participant (207) on a lane (223) and / or a lane (225) is located (111) by means of the prediction module (209); Identifying (113) a lane segment (231) and / or a lane segment (233) of a lane (223) and / or a lane (225) on which the traffic participant (207) is located, wherein the lane segment (231) and / or the lane segment (233) define a sub-area of ​​the lane (223) and / or the lane (225), wherein the lane segment (231) and / or the lane segment (233) pass through delimited by lane boundary elements (221) of the lane (223) and / or lane boundary elements of the lane (225), wherein the lane segment (231) can be connected to another lane segment (231) and / or the lane segment (233) can be connected to another lane segment (233), wherein a plurality of lane segments (231) delineate the lane (223) and / or a plurality of lane segments (233) delineate the lane (225); Taking into account current traffic regulations and the course of the lane (223) and / or lane (225), the lane segment (231) and / or lane segment (233) are connected (115) to other lane segments (231) and / or lane segments (233) that are respectively connectable to one another to form the anchoring path (229).

4. The method (100) according to claim 3, wherein: The anchoring path (229) includes a lane change and / or a lane change if the lane change and / or the lane change is compatible with the course of the lane (223) and / or the lane (225) and with the prevailing traffic regulations.

5. The method (100) according to any one of claims 2 to 4, further comprising: Based on environmental sensor data (235) of at least one environmental sensor (237) of the vehicle (201), the other traffic participants (207) are classified (117) into one of the following list: vehicle, bus, truck, tram, motorcyclist, cyclist, pedestrian, animal by means of an environmental detection device; wherein an anchor path (229) is generated only for the following traffic participants (207): the traffic participants are classified as vehicle, bus, truck, tram, motorcyclist.

6. The method (100) according to any one of the preceding claims, further comprising: receiving (119) further map data (253) of a map representation (219) of an environment of the host vehicle (201) by means of the prediction module (209), wherein the further map data (253) are in the form of image data; Lane characteristics (239) of a lane on which the traffic participant (207) is traveling are read (121) from the further map data (253) by the prediction module (209); wherein the movement trajectory (205) is predicted based on trajectory data (203), map data (217) and / or an interactive graphical representation (227) of the traffic participant (207) and / or the lane characteristics (239).

7. The method (100) according to claim 6, wherein: The lane features (239) include: stop lines, traffic signs, traffic lights, crosswalks, lane environments, sidewalks, and parking lots.

8. The method (100) according to any one of the preceding claims, wherein: The arrangement relationship of the edges (215) of the interactive graphical representation (227) includes lane information about the traffic participant (207), wherein the lane information includes: The road user (207) is located in the same lane (225) as another road user (207) and / or the own vehicle (201); and / or The road user (207) is positioned on a lane (225) that is located next to a lane (225) of another road user (207) and / or of the own vehicle (201); and / or A road user (207) is located on a lane (225) that intersects a lane (225) of another road user (207) and / or of the own vehicle (201), The motion trajectory (205) is predicted based on trajectory data (203), map data (217), an interactive graphic representation (227) of the traffic participant (207) and / or the lane features (239) and / or the arrangement relationship of the interactive graphic representation (227).

9. The method (100) according to claim 8, wherein: The prediction (107) of the motion trajectory (205) includes: Based on the position information and the lane information of the interactive graphic representation (227), for a traffic participant (207) in the surroundings of the host vehicle (201), a distance (241) of the traffic participant (207) from another traffic participant (207) and / or from the host vehicle (201) is determined (123); Based on the distance (241) and speed information of the road users (207), determining (125) a probability value of a collision between the road users or with the vehicle (201); In this case, the movement trajectory is completed taking into account the determined probability value (205).

10. The method (100) according to any one of the preceding claims, wherein: The nodes (213) of the interactive graphical representation (227) include information about the type of the corresponding traffic participant (207), wherein the type is defined as one of the following list: vehicle, bus, truck, motorcyclist, cyclist, pedestrian, animal.

11. The method (100) according to any one of the preceding claims, wherein: The prediction module (209) is constructed as an artificial intelligence that is trained to implement the above-mentioned method steps, wherein the prediction module (209) includes a participant level (243), a map level (245), a fusion level (247), a merging level (249) and at least one encoder element (251), wherein the participant level (243) is configured to process the trajectory data (203) and / or generate an interactive graphical representation (227), wherein the map level (245) is configured to process the map data (217), and wherein the fusion level (247) is set up for fusing the results of the participant level (243) and the map level (245), wherein the at least one encoder element (251) is set up for processing further map data (217) constructed as image data and extracting the lane characteristics (239) and / or lane characteristics from the map data (217), wherein the merged level (249) is set up for predicting and outputting at least one movement trajectory of at least one traffic participant (207) based on the results of the fusion layer (247) and the results of the encoder element (251).

12. The method (100) according to claim 11, wherein: The artificial intelligence comprises at least one graphical neural network, and / or wherein the encoder element (251) is designed as an autoencoder and is configured to generate a latent space representation in a latent space of the lane features (239) based on image data of the further map data (217).

13. A method (300) for controlling a host vehicle (201), the method comprising: Carrying out (301) trajectory prediction according to the method (100) for trajectory prediction according to any one of claims 1 to 12; At least one control function of the host vehicle (201) is implemented (303) taking into account the predicted driving trajectory of the vehicle.

14. A computing unit (225) configured to carry out the method (100) for trajectory prediction according to any one of claims 1 to 12 and / or the method (300) for controlling a host vehicle (201) according to claim 13.

15. A computer program product (400), comprising instructions which, when the program is executed by a data processing unit, cause the data processing unit to implement the method (100) for trajectory prediction according to any one of claims 1 to 12 and / or the method (300) for controlling the vehicle (201) according to claim 13.

Images