DE102023206336A1 - method for operating a motor vehicle - Google Patents

Abstract

The invention relates to a method for operating a motor vehicle (2) with a sensor (4) for detecting its surroundings and/or a further vehicle (6), wherein information (I1 to I11), in particular a parameter, about the lane (10) used by the further vehicle (6) and/or about the further vehicle (6) is determined on the basis of measurement data (D) of the sensor (4), wherein a probability distribution (P) for a node, in particular a parent node, of a static Bayesian network is determined on the basis of the information, and wherein a probability for a turning maneuver of the further vehicle (6) is determined on the basis of the static Bayesian network (12).

Description

The invention relates to a method for operating a motor vehicle which has a sensor for detecting its surroundings and/or another vehicle.

Methods for operating a vehicle are known in which it is determined whether another vehicle is performing a turning maneuver. For example, DE 10 2017 100 029 A1 a system and method is known for predicting whether a driver of a host vehicle or a remote vehicle intends to turn left or right at an intersection or to proceed straight through an intersection before the host vehicle or the remote vehicle reaches the intersection based on a probabilistic model applying a dynamic Bayesian network. The method includes obtaining a plurality of environmental factors indicative of external parameters at or around the intersection, the environmental factors including the position and speed of the remote vehicle, and obtaining a plurality of host vehicle factors defining the operation of the host vehicle. The method then predicts the turn intention of the host vehicle and/or remote vehicle at the intersection using the model based on both the external factors and the vehicle factors by means of the model. The model may utilize learned information about previous turning maneuvers of the driver at the intersection.

Furthermore, in the DE 10 2019 132 006 A1 discloses a method for detecting a turning object. This includes detecting a vehicle driving ahead using an environmental sensor system of an ego vehicle, determining a turning probability for the vehicle driving ahead, and marking the vehicle driving ahead as a turning object if the turning probability is equal to or greater than a predetermined threshold.

If a driver assistance system, such as an adaptive cruise control (ACC), does not classify a vehicle turning or classifies it only relatively unreliably, or if such information is not provided to the driver assistance system, there is a risk that a control system, in particular a speed control system, of the ego vehicle based on the driver assistance system will not react to such a maneuver at all or will react in an uncomfortable way for the driver.

For example, with an adaptive cruise control (ACC, “active cruise control”), the ego vehicle is braked until the vehicle in front has completely completed the turn. This means that a comparatively long braking process takes place, which is followed by a correspondingly longer and/or stronger acceleration to the target speed.

The invention is based on the object of detecting a turning of another vehicle comparatively reliably and/or predicting such a turning process with comparatively little effort.

This object is achieved according to the invention by a method having the features of claim 1. Furthermore, the object is achieved according to the invention by a motor vehicle having the features of claim 10, by a computer program having the features of claim 11 and by a computer-readable medium having the features of claim 12. Advantageous embodiments and further developments are the subject of the subclaims. The statements in connection with the method also apply mutatis mutandis to the motor vehicle, to the computer program and to the computer-readable medium, and vice versa.

The method is used to operate a motor vehicle which has a sensor which is provided and set up to detect the surroundings of the motor vehicle and/or of another vehicle, in particular a vehicle driving ahead.

According to the method, the environment and/or other road users, in particular the other vehicle, are first recorded using the sensor and corresponding measurement data are provided as a result of this recording. Information about the other vehicle and/or the lane that the other vehicle uses is determined using the measurement data from the sensor. For example, the information forms a parameter that represents a property of the lane and/or the other vehicle. The information is preferably a numerical value or is represented using one.

For example, the measurement data are evaluated directly, in particular using an evaluation unit in the motor vehicle, with the information being the result of the evaluation.

For example, a (virtual) environment model is first created using the measurement data, also known as sensor data. This environment model represents the environment of the motor vehicle and recorded road users. For example, as part of the creation of the environment model, a coordinate system is defined in which the motor vehicle, its environment, in particular recorded lanes, recorded road users and/or recorded objects, such as obstacles or traffic signs, are arranged. The information is then derived from the environment model or calculated using it.

Subsequently, the information is used to determine a probability distribution, in particular an unconditional probability distribution, for a node of a static Bayesian network.

For example, the information is assigned an (event value) value between 0 and 1 as the probability of an event occurring, in particular of another vehicle turning, using a mathematical function, a table or a characteristic curve. This information is also assigned the probability that this event will not occur. This is determined accordingly as 1 minus the event value. As a result of such an assignment, a standardized probability of an event occurring and a standardized probability that this event will not occur are determined on the basis of the information.

This probability distribution is particularly useful when used as the probability distribution of a parent node, in particular a parent node that has no parents itself. The probability distribution determined according to the method thus forms an unconditional probability distribution.

Subsequently, the probability of the other vehicle turning is determined using the static Bayesian network. This includes, for example, determining a probability of whether the other vehicle turns left and/or determining a probability of whether the other vehicle turns right. Alternatively or additionally, this includes determining the probability of whether the vehicle turns, regardless of whether it turns left or right.

In summary, the method predicts a turning maneuver of the other vehicle. Using a static Bayesian network enables a comparatively reliable prediction of the turning maneuver based on the information determined and the measurement data. In comparison to a dynamic Bayesian network, in which nodes of neighboring time steps are connected to one another, the static Bayesian network only includes probability distributions for one (single) time step. Consequently, the prediction is advantageously comparatively simple.

It is expedient to make a certain probability of a turning maneuver available to an assistance function, such as an adaptive cruise control system. The assistance function then expediently sets a vehicle speed and/or a distance to the other vehicle depending on the certain probability of a turning maneuver. For example, the assistance function compares the certain probability of a turning maneuver with a predetermined threshold value and outputs a target value, in particular for the vehicle speed and/or for the distance to the other vehicle, depending on the result of the comparison.

According to a particularly suitable embodiment of the method, the probability distribution that is determined based on the information is determined using a sigmoid function with a value range between 0 and 1. In particular, the information, in particular designed as a parameter, forms an independent variable (input variable) of the sigmoid function. The sigmoid function particularly preferably comprises a parameter (function parameter, shape variable) that is predetermined and/or can be predetermined. This shape variable is expediently selected and predetermined depending on the detected environment or the other vehicle. For example, this shape variable is derived from the environment model or calculated based on it. In this way, the sigmoid function can be adapted to the environment, so that the probability of the other vehicle turning is comparatively reliable.

According to a suitable embodiment, a lateral position of the additional vehicle with respect to a center line of the lane, also referred to as the centerline, and/or the temporal derivative (change) of the lateral position is used as the information. A position of the additional vehicle in the transverse direction of the lane to the lane used by it is therefore determined and used as information. In particular, the distance between the center line of the lane to the rear position of the additional vehicle, i.e. the intersection point of the center axis of the vehicle with its rear end, is used as the lateral position.

The lateral position describes the lateral position of the other vehicle in relation to the center line of the lane, also referred to as the lane. A comparatively large lateral position and/or a positive value of the time derivative of the lateral position is an indication of a turning maneuver.

According to a suitable embodiment, a rectangle that is rotationally fixed with respect to the center line of the lane and envelops the vehicle is determined. Such a rectangle is also referred to as a "fixed rotation bounding box". This is the rectangle that is rotationally fixed with respect to the center line, i.e. non-rotatable, and that completely encloses the vehicle in a top view, with a minimal area. The area of this rectangle and/or a temporal change in this area is then determined. This area and/or its temporal change is used as the information.

If the area is larger than the area of a rectangular (non-rotationally fixed) envelope, i.e. a non-rotationally fixed bounding box, for the other vehicle, this indicates a turning maneuver. This is particularly the case if the longitudinal direction of the other vehicle is not parallel to the longitudinal direction of the lane, i.e. if a yaw angle is not equal to 0°. In comparison to using the yaw angle itself as information, a comparatively reliable, i.e. error-free prediction of the turning intention of the other vehicle is made, since both the lane geometry and the vehicle geometry are taken into account.

According to a suitable embodiment, a turning point for turning at an intersection is determined, in particular based on the environment model. The turning point used is, for example, the intersection of the center line of the lane (current lane) that the other vehicle is currently using and the center line of the (turning) lane into which the other vehicle can turn and/or which intersects with the current lane. The turning point used is, for example, the point on the center line of the current lane at which the turning lane begins in terms of the longitudinal direction of the lane.

A turning area around the turning point is then determined, in particular defined, which is expediently circular. The turning point is located within the turning area, expediently centrally in the turning area. For example, the turning area is the area that is located within a circle with a specified radius around the turning point as the center. This area is also referred to as the "region of interest".

An expected, i.e. predicted, time is then determined that the other vehicle will need to reach the turning area. In addition or as an alternative to this, the distance of the other vehicle to the turning area is determined. In addition or as an alternative to this, a temporal change in this expected time and/or this distance is also determined. This time, its temporal change, this distance and/or its temporal change are then used as the information.

Preferably, the turning area, in particular the radius used to determine it, is or will be predeterminable and/or adjustable. Consequently, the turning area is adjustable for each intersection. For example, the radius is selected proportionally to the lane width of the current lane and/or proportionally to the speed of the vehicle.

According to a suitable embodiment, a type and/or presence of a lane marking, for example a direction arrow, a solid or broken line is used as a lane boundary, in particular as the information. In addition or alternatively to this, the type and/or presence of a traffic sign and/or a traffic light position (for example "red", "green") is used as the information, in particular as the information. The type and/or presence of a lane marking, a traffic sign, and/or a traffic light position is expediently determined using the environment model or directly from the measurement data. Furthermore, the type and/or presence of a lane marking, a traffic sign, and/or a traffic light position is a truth probability distribution for a node of the Bayesian network. In particular, a table or the like is provided for this purpose, which assigns a probability distribution to the type or presence of one of these features.

According to a suitable embodiment, the information is determined as an expected, i.e. predicted (turning) time that the other vehicle needs or would need during the turning process in order to completely leave the current lane. For example, a trajectory for the turning process is predicted for the right rear end and for the left rear end of the other vehicle. The circle equation is used in particular for this purpose. Based on the current acceleration, the current speed and/or the current position of the other vehicle, a time period can be calculated for both trajectories, with the longer of the two time periods being used as the turning time period.

According to a suitable embodiment, the information used is the, in particular current, speed of the further vehicle, its acceleration, in particular longitudinal acceleration, the yaw angle and/or the temporal change of the yaw angle.

According to a suitable embodiment, a turn signal status (in particular “left turn signal on”, “right turn signal on”) of turn signals of the other vehicle is detected and this is used as information.

To determine the probability distribution for a (first) node of the Bayesian network, a predefined value, in particular 0 or 1, is assigned to the detected indicator status. For example, an associated probability distribution is stored in a memory for each possibility for the indicator status.

In addition or as an alternative to this, the distance of the motor vehicle to the other vehicle is determined as (further, second) information and this is used to determine a probability distribution for another (second) node of the Bayesian network. The first and second nodes form parents for a common (third) node. In summary, the distance to the other vehicle is taken into account to determine the probability of a turn based on the indicator status. This advantageously corresponds to a stabilization, i.e. an improvement in the reliability, of the detection of the indicator status.

A further aspect of the invention relates to a motor vehicle. This can be operated according to the method in one of the variants presented above and/or is operated according to the method. For this purpose, the vehicle comprises a sensor which is set up to detect its surroundings and/or that of another vehicle, in particular a vehicle driving ahead. The vehicle also comprises a means for carrying out the method steps of the method in one of the variants presented above. The means is expediently or comprises a computing device such as a computer and/or a processor. The motor vehicle also expediently comprises a computer-readable medium on which the Bayesian network or a calculation rule corresponding to it is stored.

A further aspect of the invention relates to a computer program. This comprises instructions which cause the motor vehicle to carry out the method steps of the method in one of the variants described above. In particular, the computer program comprises instructions which cause information about the lane used by the other vehicle and/or about the other vehicle itself to be determined, a probability distribution for a node of a static Bayesian network to be determined based on the information, and a probability for a node of the other vehicle to be determined based on the static Bayesian network. The computer program expediently also comprises instructions which cause conditional and/or unconditional probability distributions to be determined for the nodes, in particular for all nodes, of the Bayesian network. Alternatively, these probability distributions are stored in a memory and are loaded during the method.

A further aspect of the invention relates to a computer-readable medium, in particular designed as a non-volatile memory on which the computer program is stored. The computer-readable medium is, for example, a flash memory, a hard disk, a CD, a DVD, or the like. The motor vehicle preferably has the computer-readable medium. The computer-readable medium is expediently connected and/or connectable to the means for carrying out the method in terms of signal and/or data transmission technology.

• 1 anhand eines Flussdiagramms einen Verfahrensablauf zum Betrieb eines Kraftfahrzeugs, wobei anhand eines statischen Bayes'schen Netzwerks eine Abbiegewahrscheinlichkeit eines zum Kraftfahrzeug vorausfahrenden Fahrzeugs anhand von Messdaten eines Sensors des Kraftfahrzeugs bestimmt wird,
• 2 das statische Bayes'sche Netzwerk,
• 3 schematisch in Draufsicht das Fahrzeug im Bereich einer Kreuzung und dessen laterale Position,
• 4 schematisch in Draufsicht das sich einer Kreuzung nähernde Fahrzeug, welches dem Kraftfahrzeug vorausfährt, wobei ein Abbiegebereich um eine Abbiegestelle bestimmt wurde,
• 5 schematisch in Draufsicht das dem Kraftfahrzeug vorausfahrende Fahrzeug im Bereich einer Kreuzung, wobei ein das vorausfahrende Fahrzeug einhüllendes und bezüglich der Fahrspur rotationsfestes Rechteck bestimmt wurde, und
• 6 schematisch in Draufsicht das dem Kraftfahrzeug vorausfahrende Fahrzeug im Bereich einer Kreuzung, wobei zwei für ein Abbiegen des Fahrzeugs erwartete Trajektorien bestimmt wurden.

In the following, embodiments of the invention are explained in more detail with reference to a drawing. In the drawing:

• 1 using a flow chart, a process sequence for operating a motor vehicle, whereby a turning probability of a vehicle driving ahead of the motor vehicle is determined using a static Bayesian network based on measurement data from a sensor of the motor vehicle,
• 2 the static Bayesian network,
• 3 schematic top view of the vehicle in the area of an intersection and its lateral position,
• 4 schematic plan view of the vehicle approaching an intersection, which is driving ahead of the motor vehicle, whereby a turning area has been determined around a turning point,
• 5 schematically in plan view the vehicle driving ahead of the motor vehicle in the area of an intersection, whereby a rectangle enveloping the vehicle driving ahead and rotationally fixed with respect to the lane has been determined, and
• 6 schematic plan view of the vehicle preceding the motor vehicle in the area of an intersection, whereby two trajectories expected for a turning of the vehicle have been determined.

Corresponding parts and sizes are always provided with the same reference symbols in all figures.

In the 1 a flow chart is shown that represents a method for operating a motor vehicle 2. This comprises at least one sensor 4, which is provided and set up to detect the surroundings of the motor vehicle 2 and/or another vehicle 6, here for example a vehicle 6 driving ahead of the motor vehicle 2. For example, the sensor 4 is or comprises a camera, a lidar or a radar.

In the method, in a first step I, the environment and/or other road users, here the vehicle 6 driving ahead of the motor vehicle 2, are recorded using the sensor 4 and corresponding measurement data D is provided. For example, the other vehicle 6 uses the same (current) lane 10 as the motor vehicle 2; alternatively, the vehicle 6 uses an adjacent lane. For example, measurement data is provided for all recorded vehicles in accordance with the method presented here. In a second step II, based on the measurement data D, a virtual environment model 8 is calculated which represents the environment of the motor vehicle 2 and recorded road users, here the other vehicle 6. For example, as part of the formation of the environment model 8, a coordinate system is defined in which the motor vehicle 2, its environment, in particular recorded lanes, recorded road users and/or recorded objects, such as obstacles or traffic signs, are arranged and/or evaluated.

Based on this environment model 8, one or more pieces of information I ₁ to I ₁₁ about the other vehicle 6 and/or about the lane 10 used by the other vehicle 6 can be determined.

According to the example presented here, in a third step III the information I ₁ to I ₁₁ is determined using the environment model 8 and/or directly using the measurement data D. The determination of the respective information I ₁ to I ₄₁ is described below in connection with the 3 to 6 presented in comparatively detailed form.

For the information I ₁ to I _11, in a subsequent fourth step IV, either a probability distribution P _n or both a probability distribution P _n,L and a probability distribution P _n,R is determined. The determined probability distributions P _n , _{P n,L} and P _n,R are each assigned to one of the nodes K ₁ to K ₁₉ of a static Bayesian network 12, which is in the 2 The nodes K ₁ to K ₁₉ are parent nodes that do not have any parents themselves. Therefore, the probability distributions determined in step IV are unconditional probability distributions.

The respective probability distribution P _n in turn comprises two probabilities, namely Pt and P _f , where Pt represents the probability that the vehicle 6 will turn left or right, and where P _f represents the probability that the vehicle 6 will turn. In an analogous manner, the probability distribution P _n,L in turn comprises two probabilities, namely Pt and P _f , where Pt represents the probability that the vehicle 6 will turn left. , and where P _f represents the probability that the vehicle 6 will not turn left. Accordingly, the probability distribution P _n,R comprises the two probabilities Pt and P _f , where Pt represents the probability that the vehicle 6 will turn right, and where P _f represents the probability that the vehicle 6 will not turn right.

The following applies: $${\text{P}}_{\text{f}}=1-{\text{P}}_{\text{t}}.$$

In summary, the probabilities Pt and P _f form the respective probability distribution P _n or P _n,L or P _n,R , at the respective assigned node.

The information I ₁ to I ₉ is each a numerical value x or is represented by a numerical value x. The probability Pt is preferably determined as follows using a sigmoid function f _S (x) (equation (1)): $$P_t=\{\begin{array}{c}{f}_{s\left(x\right)},wenn{x}_{\text{min}}\le \text{x}\le {\text{x}}_{max}\\ \mathrm{0,}wennx<x_{min}\\ \mathrm{1,}wennx>x_{max}\end{array}$$

a, x_min, x_max sind Parameter (Formvariable). Diese sind bevorzugt vorgebbar und/oder anpassbar. Auf diese Weise ist die Sigmoidfunktion an das erfasste Umfeld anpassbar. Beispielsweise wird die jeweilige Formvariable aus dem Umfeldmodell 8 abgeleitet, bzw. anhand diesem berechnet. Alternativ hierzu sind die Parameter a, x_min, x_max fest, also unveränderlich, vordefiniert. Die Parameter sind dabei anhand von A-Priori-Wissen bestimmt. Beispielsweise werden diese anhand von Versuchsdaten oder anhand einer Simulation ermittelt oder alternativ hierzu abgeschätzt.The sigmoid function f _S (x) is defined, for example, by: $$f_s\left(x\right)=\frac{1}{1+e^{\frac{a}{{}^{\left(x_{min}-x_{max}\right)\cdot \left(x-\mathrm{0,5}(x_{min}+x_{max}\right))}}}}$$

a, x _min , x _max are parameters (form variables). These can preferably be specified and/or adapted. In this way, the sigmoid function can be adapted to the recorded environment. For example, the respective form variable is derived from the environment model 8 or calculated using it. Alternatively, the parameters a, x _min , x _max are fixed, i.e. unchangeable, predefined. The parameters are determined using a priori knowledge. For example, they are determined using test data or a simulation or, alternatively, estimated.

The parameters a, x _min , x _max are appropriately chosen for each of the information items I ₁ to I ₉ , and can therefore differ for two of the information items.

In summary, a corresponding probability distribution P _n , P _n,R , P _n,L is determined based on the respective information I ₁ to I ₉ .

In step V, a probability P _Turn for a turning maneuver of the other vehicle 6, a probability P _L for a turning maneuver to the left, and/or a probability P _R for a turning maneuver to the right are determined using the static Bayesian network 12. Furthermore, for example, the probability P _G is determined that neither a turn to the right nor to the left is made, i.e. that the vehicle 6 drives straight ahead.

In the 2 the static Bayesian network 12 is shown. This comprises the parent nodes K ₁ to K ₁₉ with unconditional probability distributions P _n P _n,R , P _n,L . In other words, one of the probability distributions P _n P _n,R , P _n,L determined in step IV is assigned to one of the nodes K ₁ to K ₁₉ .

The Bayesian network 12 also includes the nodes K ₂₁ and K ₂₂ . The nodes K ₁ , K ₃ , K ₄ , K ₆ , K ₈ , K ₁₀ , K ₁₂ , K ₁₃ , K ₁₈ and one node K ₂₀ (cf. 6 ) Parents of node K ₂₁ . Nodes K ₂ , K ₃ , K ₅ , K ₇ , K ₉ , K ₁₁ , K ₁₂ , K ₁₂ , K ₁₉ and K ₂₀ are parents of node K ₂₂ .

In summary, node K ₂₁ is connected to those nodes to which a probability distribution P _n,L or a probability distribution P _n is assigned. In addition, node K ₂₁ is connected to node K _20. Node K ₂₁ is therefore not connected to those nodes to which a probability distribution P _n,R is assigned. Accordingly, node K ₂₂ is connected to those nodes to which a probability distribution P _n,R or a probability distribution P _n is assigned. In addition, node K ₂₂ is connected to node K _20. Node K ₂₂ is therefore not connected to nodes that are assigned a probability distribution P _n,L .

Furthermore, the Bayesian network 12 comprises a node K ₂₃ , whose parents are the nodes K ₂₁ and K ₂₂ .

The respective probability distribution P _v at the nodes K ₂₀ , K ₂₁ , K ₂₂ and K ₂₃ is predefined and suitably stored in a memory of the motor vehicle 2. The respective probability distribution P _v is determined, for example, on the basis of a priori knowledge.

For example, these are determined on the basis of tests or simulations or, alternatively, estimated.

For example, for node 23, the probability distribution P _v shown in the table below is used Turn left Turn right driving straight ahead No left turns according to K ₂₁ No right turns according to K ₂₂ 0 0 1 Turn left according to K ₂₁ No right turns according to K ₂₂ 1 0 0 No left turns Turn right according to K ₂₂ 0 1 0 according to K ₂₁ Turn left according to K ₂₁ Turn right according to K ₂₂ 0.33 0.33 0.33

In the 3 the further vehicle 6 is shown in the area of an intersection of a first (current) lane 10, in which the vehicle 6 and/or the motor vehicle 2 (EGO vehicle) is driving, and a second lane 14, into which the vehicle 6 can turn. The center line of the current lane 10 is provided with the reference symbol M. Furthermore, in the 2 the lateral position p _lat of the vehicle 6 with respect to the center line M is shown, whereby the reference point of the vehicle 6 is its rear pose p _H , i.e. its central rear end in the vehicle's transverse direction. The lateral position p _lat is calculated using the environment model 8 or directly from the measurement data D. The lateral position p _lat is used here as the (first) information I _1. Based on the lateral position, both the probability distribution P _n,L for a turn to the left for the node K ₁ and the probability distribution P _n,R for a turn to the right for the node K ₂ are determined according to equation (1). The following applies: $$x=p_{lat}.$$

According to an alternative not shown further, in addition or as an alternative to this, the temporal derivative of the lateral position p _lat is used as information and from this a corresponding probability distribution for a node of the Bayesian network 12 is determined.

The vehicle 6 is traveling at a current speed v and has a current acceleration. The amount of the current speed v is used as the (second) information I ₂ . According to an alternative not shown further, in addition or as an alternative to this, the current acceleration, in particular its amount, of the vehicle 6 is used as information I ₂ or as further information and a corresponding probability distribution for the Bayesian network 12 is determined from this. Based on the amount of the current speed, the probability distribution P _n for the node K ₃ of the static Bayesian network 12 is expediently determined according to equation (1). The following applies: $$x=\left|v\right|.$$

The temporal change of the yaw angle α (yaw angle), i.e. the angle between the vehicle's longitudinal direction and the longitudinal direction of the center line M, is used as the (third) information I _3. The temporal change of the yaw angle α is used to determine both the probability distribution P _n,L for a turn to the left for the node K ₄ as well as the probability distribution P _n,R for a turn to the right for the node K ₅ are determined according to equation (1). The following applies: $$x=\dot{\alpha }.$$

In addition, the yaw angle α itself is determined as information I ₄ and based on this both the probability distribution P _n,L for a turn to the left for the node K ₆ and the probability distribution P _n,R for a turn to the right for the node K ₇ are determined according to equation (1). The following applies: $$x=\alpha .$$

In the 4 The vehicle 6 is shown schematically as it approaches an intersection. Furthermore, the 4 A turning point p _A is shown schematically. This is defined here, for example, as the point on the center line M of lane 10 at which the crossing lane 14 begins in terms of the lane's longitudinal direction. This turning point p _A is determined, for example, within the framework of the environment model 8. Furthermore, in the 4 a turning area ROI is shown, which is designed as a circular area of a circle around the turning point p _A as the center. For example, the radius of the circle is twice the width of the lane 10. For example, the radius is selected to be proportional to the current speed of the vehicle 6.

A distance d _ROI of the vehicle 6 to the turning area ROI is used as the information I ₅ . In addition, an expected time t _e that the vehicle 6 needs to reach the turning area is used as the information I ₆ . To determine the expected time, for example, the equation of motion and the known position, the current speed v and the current acceleration of the vehicle 6 are used.

In summary, the method determines the turning point p _A , the turning area ROI is determined, and the _expected time t _e until the vehicle 6 reaches the turning area ROI, or the distance d _F of the other vehicle 6 to the turning area ROI, is used as information I 5 and I ₆ .

Based on the information I ₅ , i.e. based on the distance d _ROI , both the probability distribution P _n,L for a turn to the left for the node K ₈ and the probability distribution P _n,R for a turn to the right for the node K ₉ are determined according to equation (1). The following applies: $$x=d_{ROI}.$$

In der 5 ist schematisch das weitere Fahrzeug 6 im Bereich einer Kreuzung dargestellt. Weiterhin ist in der 5 eine sogenannte „fixed rotation bounding box“ des Fahrzeugs dargestellt. Diese wird beispielsweise im Rahmen des Umfeldmodells 8 berechnet. Die fixed rotation bounding box ist also dasjenige Rechteck B mit minimaler Fläche, welches das Fahrzeug 6 vollständig einhüllt, also umfasst, wobei die Seiten des Rechtecks parallel bzw. senkrecht zur Mittellinie M sind. Weiterhin wird die Fläche A_B dieses Rechtecks ermittelt.Based on the expected time duration t _{e ,} both the probability distribution P _n,L for a left turn for node K ₁₀ and the probability distribution P _n,R for a right turn for node K ₁₁ are determined according to equation (1). The following applies: $$x=t_e.$$

In the 5 The other vehicle 6 is shown schematically in the area of an intersection. Furthermore, in the 5 a so-called "fixed rotation bounding box" of the vehicle is shown. This is calculated, for example, within the framework of the environment model 8. The fixed rotation bounding box is therefore the rectangle B with the minimum area that completely envelops the vehicle 6, i.e. encompasses it, with the sides of the rectangle being parallel or perpendicular to the center line M. The area A _B of this rectangle is also determined.

wobei W die Breite des Fahrzeugs 6, L die Länge des Fahrzeugs 6 und α der Gierwinkel ist. Die Länge L und die Breite B des Fahrzeugs 6 werden dabei beispielsweise anhand der Messdaten D bestimmt, oder, beispielsweise anhand einer Erkennung eines Typs oder eines Modells des Fahrzeugs 6, abgeschätzt.This results in: $$A_B=\left(\text{sin }\alpha \cdot L+\text{cos }\alpha \cdot W\right)\cdot \left(\text{cos }\alpha \cdot L+sin\alpha \cdot W\right),$$

where W is the width of the vehicle 6, L is the length of the vehicle 6 and α is the yaw angle. The length L and the width B of the vehicle 6 are determined, for example, based on the measurement data D or, for example, estimated based on recognition of a type or model of the vehicle 6.

In summary, the method determines the rectangle B enveloping the other vehicle 6 and its area A _B is determined as information I ₇ . Based on the area A _B, the probability distribution P _n for the node K ₁₂ of the static Bayesian network 12 is determined in accordance with equation (1). The following applies: $$x=A_B.$$

According to an alternative not further shown, in addition or as an alternative to this, the temporal change of this area A _B is determined as information I ₇ or as further information and from this a corresponding probability distribution is determined for the Bayesian network 12.

In the 6 the vehicle 6 is shown in the area of an intersection. Furthermore, two trajectories T ₁ and T ₂ are shown. These represent an expected path of the rear left end p _L or the rear right end p _R of the vehicle 6 when turning at the intersection to the lateral end of the current lane 10. For modeling, for example, a circle equation is used for the respective trajectory T ₁ or T _2. The radius is chosen in particular depending on the speed v of the vehicle 6. Using an equation of motion and the current acceleration and speed of the vehicle, a time is then determined for each of the trajectories T ₁ and T ₂ using an equation of motion, the current acceleration a and speed v of the vehicle, until the rear left end of the vehicle has covered the trajectory T ₁ and a time is determined until the rear right end of the vehicle 6 has covered the trajectory T ₂ . The longer duration of these two durations thus corresponds to the expected duration of time that the vehicle 6 needs to completely leave the current lane 10. This duration t _ab is used as information I ₈

Based on the time period t _ab, both the probability distribution P _n,L for a left turn for node K ₁₃ and the probability distribution P _n,R for a right turn for node K ₁₄ are determined according to equation (1). The following applies: $$x=t_{ab}.$$

Furthermore, a turn signal status of the turn signals of the vehicle 6 is recorded and used as (tenth) information I _10. In particular, the turn signal status is one of the following possibilities, namely “left turn signal on” and “right turn signal on”. The turn signal status is assigned a probability distribution P _n,L for a left turn for the node K _15. Pt is expediently 1 if the turn signal status is “left turn signal on”, otherwise Pt is 0. Furthermore, the turn signal status is assigned a probability distribution P _n,R for a right turn for the node K _16. Pt is expediently 1 if the turn signal status is “right turn signal on”, otherwise Pt is 0. Expediently, an associated probability distribution P _n,R and an associated probability distribution P _n,L are stored in a memory of the motor vehicle for each of the possible turn signal statuses.

Die Knoten K₁₅ und K₁₆, welche dem Blinkerstatus zugeordnet sind, und der Knoten K₁₇, welcher dem Abstand d_F zugeordnet sind, bilden jeweils einen Elternknoten den gemeinsamen Knoten K₂₀ (2). Folglich erfolgt am Knoten K₂₀ eine Verknüpfung. Dies entspricht einer Erkennung einer Blinkeraktivität unter Berücksichtigung des Abstands d_F zwischen dem Kraftfahrzeug und dem weiteren Fahrzeug 6, so dass die Gefahr eine Fehlerkennung der Blinkeraktivität reduziert ist. Der Knoten K₂₀ ist mit den Knoten K₂₁ und K₂₂ verbunden.In addition, the distance d _F of the motor vehicle 2 to the vehicle 6 is determined and this is used as the information I _9. The probability distribution P _n for the node K ₁₇ of the static Bayesian network 12 is conveniently determined using equation (1). The following applies: $$x=d_F.$$

The nodes K ₁₅ and K ₁₆ , which are assigned to the indicator status, and the node K ₁₇ , which is assigned to the distance d _F , each form a parent node, the common node K ₂₀ ( 2 ). Consequently, a connection is made at node K _20. This corresponds to a detection of a turn signal activity taking into account the distance d _F between the motor vehicle and the other vehicle 6, so that the risk of incorrect detection of the turn signal activity is reduced. Node K ₂₀ is connected to nodes K ₂₁ and K ₂₂ .

According to an alternative not shown, the nodes K ₁₅ and K ₁₇ form parent nodes for a first child node and the nodes K ₁₆ and K ₁₇ form parents for a second child node, whereby the first and the second child node are parent nodes for the node K ₂₀ .

A predefined context information detected by the sensor 4, namely the type and/or presence of a lane marking, a traffic sign or a traffic light position, is assigned Probability distribution P _n,L for the node K ₁₈ and/or P _n,R for the node K _19. The context information is therefore used as the (eleventh) information I _11. For example, a solid line as a lane marking is assigned the value 0 for the probability Pt. Furthermore, for example, the probability Pt for the probability distribution P _n,L is assigned the value 1 if the lane only has a direction arrow for turning left. For each of the possible context information, an assigned probability distribution P _n , an assigned probability distribution P _n,R , and/or an assigned probability distribution P _n,L is expediently stored in a memory of the motor vehicle 2.

As in the 4 As can also be seen, the motor vehicle 2 comprises a computer-readable medium 16, which is designed as a non-volatile memory. This is connected to the computing device 18 by means of signal and/or data transmission. The computing device is or comprises, for example, a computer or a processor. This is also connected to the sensor 4 by means of signal and/or data transmission, so that the measurement data can be transmitted from the sensor 4 to the computing unit 18. A computer program 20 is stored on the computer-readable medium 16. This includes commands that cause the motor vehicle 2 to carry out the method steps of the method in one of the variants shown above. For example, the commands, in particular when they are executed by the computing device 18, cause the environment model Um to be determined on the basis of the measurement data D, the information I ₁ to I ₁₁ and their associated probability distribution P _I,N to be determined on the basis of the environment model 8, these probability distributions to be used for the static Bayesian network 12, and the probability P _Turning , the probability P _L and/or the probability P _R to be determined on the basis of the static Bayesian network 12.

The computing device 18 thus forms or is a component of a means of the motor vehicle 2 for carrying out the method.

The invention is not limited to the embodiments described above. Rather, other variants of the invention can be derived from this within the scope of the claims by a person skilled in the art without departing from the subject matter of the invention. In particular, all individual features described in connection with the embodiments and/or in the claims can also be combined with one another in other ways without departing from the subject matter of the invention.

list of reference symbols

2

motor vehicle

4

sensor

6

vehicle

8

environment model

10

current lane/first lane

12

Bayesian network

14

second lane

16

storage / computer-readable medium

18

computing device

20

computer program

α

yaw angle

B

enveloping rectangle

D

measurement data

dF

distance of the vehicle to the motor vehicle

dROI

distance of the vehicle to the turning area

K1 to K23

node

L

length of the vehicle

M

center line of the lane

I1 to I11

information

pA

turning point

PTurn

probability of the vehicle turning

pH

rear float

plat

lateral position

PL

Probability of the vehicle turning left

pL

rear left end of the vehicle

Pn

probability distribution

Pn,L

probability distribution

Pn,R

probability distribution

PR

Probability of the vehicle turning right

pR

rear right end of the vehicle

ROI

turning area

T1

trajectory

T2

trajectory

v

speed of the vehicle

W

width of the vehicle

QUOTES INCLUDED IN THE DESCRIPTION

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

Cited patent literature

• DE 10 2017 100 029 A1 [0002]
• DE 10 2019 132 006 A1 [0003]

Claims (12)

Method for operating a motor vehicle (2) with a sensor (4) for detecting its surroundings and/or that of another vehicle (6), in particular a vehicle driving ahead (6), - wherein, based on measurement data (D) from the sensor (4), information (I ₁ to I ₁₁ ), in particular a parameter, about the lane (10) used by the other vehicle (6) and/or about the other vehicle (6) is determined, - wherein, based on the information (I ₁ to I ₁₁ ), a probability distribution (P _n , P _n,L , P _n,R ) for a node (K ₁ to K ₁₉ ), in particular a parent node, of a static Bayesian network (12) is determined, and - wherein, based on the static Bayesian network (12), a probability (P _turn , P _L , P _R ) for a turning maneuver of the other vehicle (6) is determined. procedure according to claim 1 , characterized in that the probability distribution (P _n , P _n,L , P _n,R ) is determined using a sigmoid function, preferably having a predefinable parameter. procedure according to claim 1 or 2 , characterized in that a lateral position (p _lat ) of the further vehicle with respect to a center line (M) of the lane (10) and/or the time derivative of the lateral position (p _lat ) is used as information ( _I 1 ). Method according to one of the Claims 1 until 3 , characterized in that a rectangle (B) which is rotationally fixed with respect to the center line (M) of the lane (10) and envelops the further vehicle (6) is determined, and that its area (A _B ) and/or the temporal change of this area (A _B ) is used as information (I _e ). Method according to one of the Claims 1 until 4 , characterized in that - a turning point (p _A ) is determined, - a turning area (ROI), in particular a circular one, is determined around the turning point (p _A ), and - an expected time period (t _e ) until the vehicle (6) reaches the turning area (ROI), or the distance (d _F ) of the further vehicle (6) to the turning area (ROI), is used as information ( _I 4 , I ₅ ). Method according to one of the Claims 1 until 5 , characterized in that a probability distribution for a node (K ₁₈ , K ₁₉ ) of the Bayesian network (12) is assigned to a type and/or presence of a road marking, a traffic sign, and/or a traffic light position. Method according to one of the Claims 1 until 6 , characterized in that an expected time period (t _ab ) is determined as information (I ₈ ) which the further vehicle (6) needs during a turning maneuver to completely leave the lane (10). Method according to one of the Claims 1 until 7 , characterized in that a speed (v), an acceleration (a), a yaw angle (α), and/or a temporal change in the yaw angle (α) of the further vehicle ( ₆ ) is used as information (I ₂ , I 3 ). Method according to one of the Claims 1 until 8 , characterized in that - a turn signal status of turn signals of the further vehicle (6) is used as information (I ₇ ), and/or - that based on the turn signal status a probability distribution (P _n,L , P _n,R ) is determined for one of the nodes (K ₁₅ , K ₁₆ ) of the Bayesian network (12), that based on the distance (d _F ) of the motor vehicle (2) to the further vehicle (6) a probability distribution (P _n ) for a further of the nodes (K ₁₇ ) of the Bayesian network (12) is determined, and that these nodes (K ₁₅ , K ₁₆ , K ₁₇ ) form parents for a further of the nodes (K ₂₀ ) of the Bayesian network (12). Motor vehicle (2), with a sensor (4) for detecting the surroundings and with a means (18) for carrying out the method steps of the method according to one of the Claims 1 until 7 . Computer program (20) comprising commands which cause the motor vehicle (2) of the claim 10 the procedural steps of the procedure according to one of the Claims 1 until 9 executes. Computer-readable medium (16) on which a computer program according to claim 11 is deposited.

Images