Method for predicting a trajectory of a vehicle
Patent Information
- Application Number
- EP2023776628
- Authority / Receiving Office
- EP · EP
- Patent Type
- Applications
- Current Assignee / Owner
- Priority Date
- 2022-09-27
- Filing Date
- 2023-09-22
- Publication Date
- 2025-08-06
AI Technical Summary
Current methods for predicting a vehicle's trajectory, especially for single-track vehicles like eBikes, struggle to accurately forecast longer-term paths due to reliance on driver intention, which cannot be solely determined by vehicle conditions, and often result in inaccuracies when combining kinematic models with environmental factors.
A method that involves obtaining map information, determining road networks, and using variables like linear acceleration, angular acceleration, and yaw rate to predict kinematic trajectories, combined with machine learning models to estimate probabilities of possible paths, ensuring physically plausible and accurate trajectory predictions.
Enables the prediction of both short-term and longer-term vehicle trajectories with high accuracy by integrating kinematic and environmental data, improving collision avoidance systems by providing reliable trajectory information to other road users.
Smart Images
Figure 1.1
Abstract
Description
[0001] Description
[0002] title
[0003] Method for predicting a trajectory of a vehicle
[0004] The invention relates to a method for predicting a trajectory of a vehicle, in particular a single-track vehicle such as an eBike.
[0005] The invention further relates to a vehicle, in particular a single-track vehicle such as an eBike, designed to predict a trajectory of the vehicle.
[0006] Although generally applicable to any vehicle, the present invention is explained using eBikes.
[0007] State of the art
[0008] To avoid collisions, it has become known that vehicles transmit their expected trajectory to other road users, allowing them to adjust their own trajectories accordingly. To do this, the vehicles estimate their trajectory based on factors such as the current speed or the vehicle's orientation. It is also possible for vehicles to estimate whether they are on a collision course with another road user based on their own trajectory and issue appropriate warnings to the driver.
[0009] It has also been demonstrated that the trajectory is estimated using a kinematic model combined with sensor data. This allows the vehicle's short-term trajectory to be estimated correctly with a high degree of probability. Medium-term trajectories are primarily based on the driver's intention, for example, the desired route. This cannot be determined solely from the vehicle's condition.
[0010] Disclosure of the invention
[0011] In one embodiment, the present invention provides a method for predicting a trajectory of a vehicle, in particular a single-track vehicle such as an eBike, comprising the steps:
[0012] Obtaining map information from a specified area around the current position of the vehicle,
[0013] Determine a road network based on the received map information,
[0014] Determining one or more possible paths of the vehicle based on the identified road network,
[0015] Determining a kinematic trajectory of the vehicle and / or a state of the vehicle based on at least one of the variables linear acceleration, angular acceleration, yaw rate, speed, direction, orientation, position, driver cadence, drive torque of the vehicle, driver drive torque,
[0016] Determining one or more possible trajectories of the vehicle based on the one or more determined paths, and estimating at least a probability that the one or more possible trajectories are followed by the vehicle using the determined kinematic trajectory and / or the determined state.
[0017] In one embodiment, the present invention provides a vehicle, in particular a single-track vehicle such as an eBike, designed to predict a trajectory of the vehicle, comprising: a receiving device designed to receive map information from a predeterminable area around the current position of the vehicle, a first determination device designed to determine a road network based on the received map information, a second determination device designed to determine one or more possible paths of the vehicle based on the determined road network, a first determination device designed to determine a kinematic trajectory of the vehicle and / or a state of the vehicle based on at least one of the variables linear acceleration, angular acceleration, yaw rate, speed, direction, orientation, position, driver cadence, drive torque of the vehicle, driver torque,a second determining device configured to determine one or more possible trajectories of the vehicle based on the one or more determined paths, and an estimating device configured to estimate at least one probability that the one or more possible trajectories are followed by the vehicle by means of the determined kinematic trajectory and / or the determined state.
[0018] One of the advantages achieved is that even longer vehicle trajectories can be predicted with a high degree of probability. Another advantage is that the map information can be combined with the kinematic trajectory to reliably determine trajectories.
[0019] The term "trajectory" is to be understood in the broadest sense and refers, particularly in the claims and preferably in the description, to an expected route along which the vehicle will move. In particular, a trajectory comprises a number of waypoints with associated points in time, assuming that the vehicle will be at the respective waypoints at the respective points in time.
[0020] The term "kinematic trajectory" is to be understood in the broadest sense and refers, particularly in the claims and preferably in the description, to a trajectory that depends solely on the current state of the vehicle—that is, on physical variables that describe the vehicle's driving state—such as the vehicle's speed, yaw rate, and acceleration. Environmental influences such as the course of the road on which the vehicle is traveling are not considered.
[0021] The term “linear acceleration” is to be understood in the broadest sense and refers, particularly in the claims and preferably in the description, to an acceleration parallel to the direction of travel of the vehicle.
[0022] Further features, advantages and further embodiments of the invention are described below or will become apparent thereby.
[0023] According to an advantageous development of the invention, at least one of the following variables is used to determine the one or more possible paths: speed, linear acceleration, angular acceleration, direction, yaw rate, position, drive torque, driver torque, position, and orientation of the vehicle. One advantage of this is that the determined possible paths can be determined more precisely. For example, the determined possible paths may not include tight curves if the vehicle's speed is high.
[0024] According to an advantageous development of the invention, the at least one probability is estimated based on a cross-correlation of the possible trajectories with the kinematic trajectory. A cross-correlation here is a comparison of two trajectories to identify similar trajectory paths; in particular, a correlation between the two trajectories is determined. Those trajectories that are similar to those of the kinematic trajectory are more likely to actually be driven by the vehicle. This relationship can be determined through cross-correlation, allowing the probability to be estimated more accurately.
[0025] According to an advantageous development of the invention, the one or more possible trajectories comprise the kinematic trajectory. A possible trajectory can be the kinematic trajectory, for example, if the driver is driving through a field rather than along a road. In this way, a larger number of possible trajectories can be determined. According to an advantageous development of the invention, the one or more trajectories are determined based on physical boundary conditions. Possible physical boundary conditions include, for example, a turning circle that is too small or acceleration that is too high. In this way, the trajectories can be checked for plausibility so that physically implausible trajectories are not determined or the determined trajectories are physically possible.
[0026] According to an advantageous development of the invention, the one or more trajectories are determined based on the kinematic trajectory. This improves the accuracy of the trajectories because the current status of the vehicle is taken into account. For example, this allows the times at which the vehicle will be located at waypoints along the trajectory to be estimated.
[0027] According to an advantageous development of the invention, the one or more paths are determined using parametrically modeled curves, in particular Bézier curves. Roads can have sharp corners, bends, or edges in their course, so that the road network also has corresponding corners, bends, or edges. Thus, determined paths within the road network could also have corners or the like, for example, when a path is determined where the vehicle turns. To smooth these - unrealistic - paths, Bézier curves can be used so that the determined paths more realistically reflect the actual course. As a result, the determined paths exhibit a smaller deviation from the physically possible trajectories of the vehicle.
[0028] According to an advantageous development of the invention, the one or more trajectories are determined using a machine learning model and / or the probabilities are estimated using the machine learning model. Medium-term trajectories, in particular, depend particularly on the driver's intentions. The driver's intention may depend on the road network; for example, they may turn at an intersection. To do so, they would brake the vehicle and possibly already slightly change the vehicle's direction. Thus, there is a connection between the road network, the current state of the vehicle, and the expected trajectory. Using a machine learning model, this connection can be recognized, and the machine learning model can be trained to calculate expected trajectories based on the road network and the vehicle's state.The machine learning model can calculate both the trajectories and the associated probabilities. One advantage of this is that trajectories can be determined reliably.
[0029] According to an advantageous development of the invention, one or more possible paths and / or a state of the vehicle are provided to the machine learning model as input variables. Based on the possible paths determined by the road network and the state of the vehicle, for example, the speed, possible trajectories and associated probabilities can be determined using the machine learning model. This allows the machine learning model to effectively determine possible trajectories. It is also possible for the kinematic trajectory to be provided to the machine learning model as input variables, either in addition to or as an alternative to the state of the vehicle.
[0030] According to an advantageous development of the invention, an acceleration and / or yaw rate profile is calculated using the machine learning model. The machine learning model can have an acceleration and / or yaw rate profile as an output variable. An acceleration and / or yaw rate profile contains expected accelerations and / or yaw rates of the vehicle at several future points in time. Based on these variables, possible trajectories can be determined. One advantage of this is that the possible trajectories can be determined easily.
[0031] According to an advantageous development of the invention, the one or more trajectories are determined based on the acceleration and / or yaw rate profile and a kinematic model. The kinematic model can, for example, incorporate physical limits so that the possible trajectories are realistically drivable. Furthermore, an initial yaw rate can be used to calculate the possible trajectories. Further important features and advantages of the invention emerge from the subclaims, the drawings, and the associated description of the figures. It is understood that the features mentioned above and those to be explained below can be used not only in the respective combination specified, but also in other combinations or on their own, without departing from the scope of the present invention.
[0032] Preferred embodiments and embodiments of the present invention are illustrated in the drawings and are explained in more detail in the following description.
[0033] It shows in schematic form
[0034] Figure 1 shows steps of a method according to an embodiment of the present invention;
[0035] Figure 2 A road network with trajectories according to an embodiment of the present invention,
[0036] Figure 3 A system of a machine learning model according to an embodiment of the present invention, and
[0037] Figure 4 A vehicle according to an embodiment of the present invention.
[0038] Figure 1 shows in schematic form steps of a method according to an embodiment of the present invention.
[0039] Figure 1 describes steps of a method for predicting a vehicle's trajectory. In a first step S1, the vehicle receives map information from a predeterminable area around the vehicle's current position. The map information can, for example, be stored in the vehicle or provided via a mobile communications interface. Since the vehicle's trajectory is only predicted for a few seconds, a local section of the map around the vehicle's current position can be used, for example, a radius of 50 meters. The vehicle's current position can be determined, for example, using a GPS receiver.
[0040] The map information may include additional irrelevant data, such as marked buildings. Therefore, in a second step S2, a road network is determined based on the received map information. The road network includes, in particular, information about the route of the roads in the vehicle's surroundings, and possibly the width of the roads. Based on the road network, one or more possible paths for the vehicle are determined in a third step S3. The paths describe possible routes that the vehicle can take. For example, the vehicle could turn at an intersection or continue straight ahead. The state of the vehicle can be used, for example, to determine the paths. For example, in a curve, the radius of the trajectory can be increased if the vehicle is traveling at high speed. Roads usually meet at an angle.Accordingly, paths determined in the road network could also be angled. However, such an angled path is not actually drivable by a vehicle. Therefore, the paths can be determined using Bezier curves, so that the paths do not exhibit sharp angles.
[0041] In a fourth step S4, a kinematic trajectory of the vehicle and / or a state of the vehicle is determined based on at least one of the following variables: linear acceleration, angular acceleration, yaw rate, speed, direction, orientation, position, driver cadence, vehicle drive torque, or driver drive torque. The kinematic trajectory corresponds to the trajectory the vehicle would travel if it continued driving according to its current state, for example, at the same speed or in the same direction.
[0042] In a fifth step S5, one or more possible trajectories of the vehicle are determined based on the one or more determined paths. To determine the possible trajectories, the determined paths can be combined with the kinematic trajectory. In this way, it is possible to determine at which times the vehicle will be located at waypoints along the path. It is also possible to adapt the paths based on the kinematic trajectory in order to obtain the possible trajectories. In addition, the trajectories can be adapted with regard to physical restrictions, such as a maximum speed in a curve. It is also possible to determine the possible trajectories using a machine learning model. In this way, a group of possible trajectories is obtained, with the kinematic trajectory also being part of the group.For example, the possible paths and / or the possible trajectories can be determined using a Kalman filter. In a sixth step S6, at least one probability that the vehicle follows the one or more possible trajectories is estimated using the determined kinematic trajectory and / or the determined state. By cross-correlating the possible trajectories with the kinematic trajectory or using the machine learning model, the probabilities that the vehicle travels along a specific trajectory can be estimated.
[0043] Figure 2 shows in schematic form a road network with trajectories according to an embodiment of the present invention.
[0044] The road network 201 indicates the course of the roads. Starting from the vehicle's starting point 202, three trajectories are determined. The two possible trajectories 203, 203' are determined using Bezier curves and the vehicle's state, such as the vehicle's speed, yaw rate, position, and direction, with the vehicle's starting point being the starting point of the Bezier curves. The kinematic trajectory 204 shows the expected course of the vehicle if the road network is ignored. By correlating the kinematic trajectory 204 with the possible trajectories 203, 203', the probabilities that the vehicle will follow the possible trajectories 203, 203' can be calculated. The actual trajectory 205 of the vehicle is similar to the possible trajectory 203. Thus, the vehicle has approximately followed one of the possible trajectories 203, 203'.
[0045] Figure 3 shows in schematic form a system of a machine learning model according to an embodiment of the present invention.
[0046] The system's input 301 includes a set of possible paths a vehicle can take and the vehicle's state. The possible paths can be determined, for example, according to step S3 of Figure 1, and the state can be determined, for example, according to step S4 of Figure 1. A kinematic trajectory can also be used. Using a path encoder 302, the possible paths are coded so that they can be used as input for the neural network 304. Similarly, the vehicle's state is converted into an input for the neural network 304 by a state encoder 303.
[0047] The neural network 304 calculates a future acceleration and / or yaw rate profile and / or yaw rate deviation profile 305. Based on this, a group of possible trajectories 307 is calculated using a kinematic model 306, taking into account an initial yaw rate 309 and the possible paths. The kinematic model 306 includes, in particular, physical boundary conditions of the vehicle, for example, a maximum speed or a minimum turning circle, so that the group of calculated trajectories 307 is realistic. Furthermore, the kinematic model 306 can reduce the deviation of the group of trajectories 306 from the possible paths 310. The neural network 304 can have been trained based on routes traveled.
[0048] The neural network 304 can further calculate a group of probabilities 308 that the vehicle travels along the respective trajectories.
[0049] Figure 4 shows in schematic form a vehicle according to an embodiment of the present invention.
[0050] The vehicle 1, here in the form of an eBike, is designed to predict a trajectory of the vehicle 1. For this purpose, the vehicle 1 comprises: a receiving device 2, designed to receive map information from a predeterminable area around the current position of the vehicle 1, a first determination device 3, designed to determine a road network based on the received map information, a second determination device 4, designed to determine one or more possible paths of the vehicle 1 based on the determined road network, a first determination device 5, designed to determine a kinematic trajectory of the vehicle 1 and / or a state of the vehicle 1 based on at least one of the variables linear acceleration, angular acceleration, yaw rate, speed, direction, orientation, position, driver cadence, drive torque of the vehicle 1, driver torque, a second determination device 6,designed to determine one or more possible trajectories of the vehicle 1 based on the one or more determined paths, and an estimator 7 designed to estimate probabilities that the one or more possible trajectories are followed by the vehicle 1 by means of the kinematic trajectory and / or the state.
[0051] The vehicle 1 is particularly designed to carry out steps S1 to S6 according to Figure 1.
[0052] Although the present invention has been described using preferred embodiments, it is not limited thereto, but can be modified in many ways
Claims
Claims 1 . Method for predicting a trajectory of a vehicle (1), in particular a single-track vehicle (1) such as an eBike, comprising the steps: Obtaining (S1) map information from a predeterminable area around the current position of the vehicle (1), Determining (S2) a road network (201 based on the received map information, Determining (S3) one or more possible paths of the vehicle (1) based on the determined road network, Determining (S4) a kinematic trajectory (204) of the vehicle (1) and / or a state of the vehicle (1) based on at least one of the variables linear acceleration, angular acceleration, yaw rate, speed, direction, orientation, position, driver cadence and / or drive torque of the vehicle (1), driver drive torque, Determining (S5) one or more possible trajectories (203, 203') of the vehicle (1) based on the one or more determined paths, and Estimating (S6) at least one probability that the one or more possible trajectories (203, 203') are followed by the vehicle (1) by means of the determined kinematic trajectory (204) and / or the determined state.
2. Method according to claim 1, wherein at least one of the variables speed, linear acceleration, angular acceleration, direction, yaw rate, position, drive torque, driver torque, position, orientation of the vehicle (1) is used to determine the one or more possible paths. Method according to one of claims 1-2, wherein the at least one probability is estimated based on a cross-correlation of the possible trajectories (203, 203') with the kinematic trajectory (204). Method according to one of claims 1-3, wherein the one or more possible trajectories (203, 203') comprise the kinematic trajectory (204). Method according to one of claims 1-4, wherein the one or more trajectories are determined based on physical boundary conditions. Method according to one of claims 3-5, wherein the one or more trajectories are determined based on the kinematic trajectory (204). Method according to one of claims 1-6, wherein the one or more paths are determined using parametrically modeled curves, in particular Bezier curves.Method according to one of claims 1 to 7, wherein the one or more trajectories are determined using a machine learning model (304) and / or the probabilities are estimated using the machine learning model (304). Method according to claim 8, wherein the one or more possible paths and / or a state of the vehicle (1) are provided to the machine learning model (304) as an input variable. Method according to one of claims 8-9, wherein an acceleration and / or yaw rate profile is calculated using the machine learning model (304). Method according to claim 10, wherein the one or more trajectories are determined based on the acceleration and / or yaw rate profile (305) and a kinematic model (306). Vehicle (1), in particular a single-track vehicle (1) such as an e-bike, configured to predict a trajectory of the vehicle (1), comprising: a receiving device (2) configured to receive map information from a predeterminable area around the current position of the vehicle (1), a first determination device (3) configured to determine a road network (201) based on the received map information, a second determination device (4) configured to determine one or more possible paths of the vehicle (1) based on the determined road network, a first determination device (5),designed to determine a kinematic trajectory (204) of the vehicle (1) and / or a state of the vehicle (1) based on at least one of the variables linear acceleration, angular acceleration, yaw rate, speed, direction, orientation, position, driver cadence, drive torque of the vehicle (1), driver torque, a second determination device (6) designed to determine one or more possible trajectories (203, 203') of the vehicle (1) based on the one or more determined paths, and an estimation device (7) designed to estimate at least one probability that the one or more possible trajectories (203, 203') are followed by the vehicle (1) using the determined kinematic trajectory (204) and / or the determined state.