DE102020214022A1 - Method for automatically executing a driving function in a vehicle - Google Patents

Abstract

A method 300 for automatically executing a driving function in a vehicle 100 and a vehicle 100 with an automatic driving function that is controlled using the method are proposed. The method 300 provides for discrete situational data relating to the driving function to be recorded 305. The discrete situational data are converted into a time-consistent smooth target path 170 by target path parameters 145, 165 being determined in two stages 310, 315. In a first step 310 Desired path parameters for position and orientation 145 are determined from the discrete situation data using a first optimization problem. In a second step 315, a curvature course of the target path 165, 170 is determined from the discrete situational data with a second optimization problem, taking into account the target path parameters for position and orientation 145. An automatic execution of the driving function is carried out 330 based on the determined target path 170.

Description

The invention relates to a method for automatically executing a driving function in a vehicle. Furthermore, the invention relates to a vehicle with an automatic driving function that is controlled according to the proposed method.

State of the art

In modern vehicles, automatic driving functions are increasingly being offered to relieve the driver. For example, such an automatic driving function can represent a highway or parking assistant, or piloted functions are also conceivable in the future. In order to determine a current relative position and alignment of the vehicle in this context, in addition to using inertial data or stationary data (e.g. map data), observations from the moving vehicle (e.g. using a camera, radar, lidar, sensor, etc.) .) can be used.

Especially in the case of driver assistance functions for multi-lane roads, the driving function or the driving task with regard to lateral guidance (in this case a target path of the vehicle movement) changes from inertial observation only in special cases (e.g. when changing lanes or in the case of a motorway exit, etc.). Even in these special cases, as a rule only the course of the desired position of the vehicle changes, but not a curvature behavior or a curvature course of the desired path. One way to represent the driving function or task with regard to lateral guidance is to use discrete data. Typically, the data is presented with respect to the moving observer. Knowledge of spatial and/or temporal derivations along the target path, which cannot be easily determined from discrete data, is a basic requirement for lateral control of the vehicle.

From the DE 10 2018 206 018 A1 a method for driving a motor vehicle and a corresponding motor vehicle are known. Based on an already specified target path for the motor vehicle, a preliminary control variable for lateral guidance of the motor vehicle and at least one piece of motor vehicle information relating to the current position and/or orientation of the motor vehicle are determined using a first vehicle model. A final control variable is determined using a second vehicle model that is designed differently from the first vehicle model. The vehicle models represent the variable to be controlled for driving the motor vehicle.

Disclosure of Invention

It is the object of the present invention to specify an improved method for automatically executing a driving function in a vehicle and an optimized vehicle.

This object is solved by the features of the independent claim. Further advantageous embodiments of the invention are specified in the dependent claims.

A method for automatically executing a driving function in a vehicle and a vehicle with an automatic driving function controlled according to the method are proposed. The method provides for capturing discrete situational data relating to the driving function. The discrete situation data can be in the form of movement data, for example, but can alternatively also be independent of a movement of the vehicle. The discrete situational data are converted into a time-consistent, smooth desired path by determining desired path parameters in two stages. In a first step, target path parameters for position and orientation are determined from the discrete situation data using a first optimization problem.

In a second step, a curvature course of the target path is determined from the discrete situation data with a second optimization problem, taking into account the target path parameters for position and orientation. An automatic execution of the driving function is carried out based on the determined target path.

The proposed method and the associated device improve safety when the vehicle function is carried out automatically. Because the two-stage determination of target path parameters, with which a time-consistent smooth target path can be described, dependencies on the movement of the vehicle can be taken into account individually in the individual steps. In particular, the course of curvature of the target path is not influenced by the movement of the vehicle, while the target path parameters for position and orientation are very strongly influenced by the movement of the vehicle. Consequently, the determination of the target path in one step can lead to a falsification of the target path parameters position and orientation and thus to a worse starting point for controlling the execution of the driving function. In concrete terms, this can therefore have a negative effect on various vehicle devices of the vehicle (for example a front-wheel steering system and/or a rear-wheel steering system), which are required for the execution of the lateral guidance, for example due to errors liable regulation and/or faulty control.

The method also efficiently determines a mathematically optimal solution. An example of how the method can be used is in lane line detection, in which discrete camera data are converted into smooth lanes or curves, so that the target lane can be differentiated. The method can generally be used when a path or a curve (in the mathematical sense) that is described with discrete data is to be described or approximately described by a continuous, smooth, parameter-dependent representation, which is necessary for the process of controlling an automatic driving function is relevant. The course of the target path over time (including its local derivatives) can be smoothed in an adjustable manner. This represents a considerable simplification for the downstream controller structure for lateral control. Furthermore, the robustness against inaccuracies and temporal inconsistencies in the discrete data can be significantly increased using the proposed method.

In a further embodiment, situation data about a first observation horizon are taken into account in the first step and situation data about a second observation horizon are taken into account in the second step. The first observation horizon is shorter than the second observation horizon. For example, the first observation horizon corresponds to a close range or local surrounding area of the vehicle and the second observation horizon to a far range. Due to the different observation horizons for the data used, the accuracy and reliability when determining the target path parameters and the target path can be improved. If the observation horizons for the two steps are selected independently of one another, then the respectively determined discrete situational data for the short observation horizon, ie the close range, can generally be determined more reliably using sensors, since this data changes dynamically. While other curve parameters such as curvature or change in curvature change less dynamically and observation data from the long-distance range are therefore better suited. Alternatively, temporal observation horizons are also conceivable.

Um den Vektor p_pose zu bestimmen, wird das integrale Optimierungsproblem $$\underset{p_{pose}}{min}{\displaystyle \sum _i{\displaystyle \underset{x_i}{\overset{x_{i+1}}{\int }}{\left({\Delta }_{i,pose}\left(p_{pose},x\right)\right)}^{2}dx}}$$

gelöst. Zur Lösung des Verfahrens können Ansätze aus dem Gebiet der Ausgleichsrechnung (z.B. die Methode der kleinsten Quadrate) mit einem dynamischen Optimierungsproblem kombiniert werden, um die optimalen Sollbahnparameter bzw. die optimale Sollbahn zu erhalten. Auch bei großen Datenmengen, die von der Sensorik, etc. erfasst werden, kann das vorgeschlagene Verfahren effizient eine optimale Lösung ermitteln, mit deren Ergebnis die Vorradlenkung bzw. Hinterradlenkung des Fahrzeugs für eine Querbewegung gesteuert werden kann. Ist die Beschreibung der Sollbahn linear in den Koeffizienten, liegt ein quadratisches Programm vor, das zum Beispiel effizient durch Matrizenalgebra gelöst werden kann.In a further embodiment, a vector p _pose summarizes the target trajectory parameters for position and orientation. A deviation between a segment of a polyline m _i x +b _i and an approximating curve f _pose (p _pose , x). $${\Delta }_{i,pOse}\left(x\right)=f_{pOse}\left(p_{pOse,}x\right)-m_{i}x+b_i$$

To determine the vector p _pose , the integral optimization problem $$\underset{p_{pOse}}{min}{\displaystyle \sum _i{\displaystyle \underset{x_i}{\overset{x_{i+1}}{\int }}{\left({\Delta }_{i,pOse}\left(p_{pOse},x\right)\right)}^2\mathrm{i.e}x}}$$

solved. To solve the method, approaches from the field of regression calculation (eg the method of least squares) can be combined with a dynamic optimization problem in order to obtain the optimal target path parameters or the optimal target path. Even with large amounts of data that are recorded by the sensors, etc., the proposed method can efficiently determine an optimal solution, with the result of which the front-wheel steering or rear-wheel steering of the vehicle can be controlled for a transverse movement. If the description of the target trajectory is linear in the coefficients, there is a quadratic program that can be efficiently solved using matrix algebra, for example.

Um den Vektor p zu bestimmen, wird das integrale Optimierungsproblem $$\underset{p}{min}{\displaystyle \sum _i{\displaystyle \underset{x_i}{\overset{x_{i+1}}{\int }}{\left({\Delta }_i\left(p,x\right)\right)}^{2}dx}}$$

gelöst. Die im vorherigen Absatz erläuterten technischen Vorteile gelten uneingeschränkt auch für diese Ausführungsform, da ebenfalls Ansätze aus dem Gebiet der Ausgleichsrechnung mit einem dynamischen Optimierungsproblem kombiniert werden. Insbesondere kann durch die zumindest teilweise Berücksichtigung der Resultate aus dem ersten Schritt, die Sollbahn exakter bestimmt werden.In a further embodiment, a vector p summarizes the nominal path parameters for the course of curvature, which contains the vector p _pose . A deviation between a segment of the polyline m _i x +b _i and an approximating curve f(p,x). $${\Delta }_i\left(x\right)=f\left(p,x\right)-\left(m_{i}x+b_i\right).$$

To determine the vector p, the integral optimization problem $$\underset{p}{min}{\displaystyle \sum _i{\displaystyle \underset{x_i}{\overset{x_{i+1}}{\int }}{\left({\Delta }_i\left(p,x\right)\right)}^2\mathrm{i.e}x}}$$

solved. The technical advantages explained in the previous paragraph also apply without restriction to this embodiment, since approaches from the field of adjustment calculation are also combined with a dynamic optimization problem. In particular, the desired path can be determined more precisely by at least partially taking into account the results from the first step.

In a further embodiment, at least one optimization problem is supplemented by at least one regularization term, which is derived from a previous determination of the target path parameters and/or from other situation data. This makes it possible to improve the temporal consistency of the determined nominal path parameters and the nominal path and to reduce the influence of signal noise. The course of the target path over time (including its local derivatives) can be smoothed in an adjustable manner. This represents a considerable simplification for the downstream controller structure for lateral control the control or activation of the vehicle equipment can be improved. Furthermore, in spite of the smoothed course over time, each path is inherently consistent with the discrete situation data, ie there is no systematic phase delay such as can occur when using a PT1 filter, for example.

In another embodiment, the regularization term is weighted.

With the help of the weighting of the regularization term, its influence can be specifically varied. Increased flexibility and robustness, as well as improved controllability of the vehicle devices can thus be ensured.

wobei λ die Gewichtungsfaktoren sind. Durch den hinzugefügten Regularisierungsterm bleibt die mathematische Struktur des Optimierungsproblems unverändert. Der Regularisierungsterm verbessert die zeitliche Konsistenz der ermittelten Sollbahn und reduziert den Einfluss von Signalrauschen. Beispielsweise kann die Schätzung mithilfe eines Kalman Filters umgesetzt sein und gegebenenfalls Situationsdaten bzw. deren Integration berücksichtigen. Eine Schätzung von Parameterwerten zu Null ist ebenfalls denkbar.In another embodiment, the regularization term includes an estimated vector pest, where the vector pest can form a parameter vector. A deviation between the estimated vector and the vector to be optimized is taken into account as an additional term in the integral optimization problem in the second step $$\underset{p}{min}{\displaystyle \sum _i{\displaystyle \underset{x_i}{\overset{x_{i+1}}{\int }}{\left({\Delta }_i\left(p,x\right)\right)}^2\mathrm{i.e}x}+{\lambda }^T{\left(p-p_{est}\right)}^2}$$

where λ are the weighting factors. The mathematical structure of the optimization problem remains unchanged by the added regularization term. The regularization term improves the temporal consistency of the determined target path and reduces the influence of signal noise. For example, the estimate can be implemented using a Kalman filter and, if necessary, take situation data or their integration into account. An estimation of parameter values to zero is also conceivable.

In a further embodiment, the automatic driving function is a lateral guidance of the vehicle. The lateral guidance of the vehicle along the target path can be carried out, for example, as part of a motorway assistant. It is also conceivable that the automatic driving function also includes longitudinal guidance or combined longitudinal and lateral guidance of the vehicle, for example as part of a piloted assistance function for automatically changing lanes.

The advantageous configurations and developments of the invention explained above and/or reproduced in the subclaims can be used individually or in any combination with one another, except, for example, in cases of clear dependencies or incompatible alternatives.

• 1 a eine schematische Darstellung einer Ausführungsform eines Fahrzeugs;
• 1 b eine schematische Darstellung einer Sollbahn;
• 2 a eine schematische Darstellung einer ersten Ausführungsform eines vorgeschlagenen Verfahrens zum automatischen Ausführen einer Fahrfunktion in einem Fahrzeug nach 1 a;
• 2 b eine schematische Darstellung einer zweiten Ausführungsform eines vorgeschlagenen Verfahrens zum automatischen Ausführen einer Fahrfunktion in einem Fahrzeug nach 1 a;
• 3 a und 3 b approximierende Kurven zur Beschreibung der Sollbahn; und
• 4 a und 4 b weitere approximierende Kurven für verschiedene Beobachtungsbereiche.

The characteristics, features and advantages of this invention described above, as well as the manner in which they are achieved, will become clearer and more clearly understood in connection with the following description of exemplary embodiments, which will be explained in more detail in connection with the schematic drawings. Show it:

• 1 a a schematic representation of an embodiment of a vehicle;
• 1 b a schematic representation of a target path;
• 2 a a schematic representation of a first embodiment of a proposed method for automatically executing a driving function in a vehicle 1 a ;
• 2 B a schematic representation of a second embodiment of a proposed method for automatically executing a driving function in a vehicle 1 a ;
• 3 a and 3 b approximating curves to describe the target path; and
• 4 a and 4 b further approximating curves for different observation areas.

It is pointed out that the figures are only of a schematic nature and are not true to scale. In this sense, components and elements shown in the figures may be exaggerated or minimized for better understanding. Furthermore, it is pointed out that the reference numbers in the figures have been chosen unchanged if identically designed elements and/or components are involved.

1 a FIG. 1 shows a schematic representation of an embodiment of a vehicle 100. Vehicle 100 includes a data acquisition unit 105, which can include sensors, for example, wherein the sensors can be designed in the form of local environment sensors with a limited range for observing the local vehicle environment. The local vehicle environment can correspond to a first observation area, ie the close range of vehicle 100. The sensors can be camera, radar, lidar, ultrasound or video-based and determine discrete situational data of moving vehicle 100. The situation data can be movement-dependent. Alternatively, the situation data can also be formed independently of the movement of vehicle 100 . In addition, the sensors can be designed as vehicle dynamics sensors, ie as Wheel speed, yaw rate or acceleration sensors. Furthermore, the data acquisition unit 105 can be coupled to an inertial navigation system, for example a global navigation satellite system, in order to acquire dynamic-independent map data in addition to the movement-dependent discrete measurement data using the inertial navigation system.

The recorded discrete situation data, which can also include stationary map data, are forwarded to a processing unit 110 . The processing unit 110 is designed, for example, to convert the discrete data into a time-consistent smooth desired path, in that the processing unit 110 determines desired path parameters in two stages. The processing unit 110 may, for example, perform a line scan or lane model estimation and based thereon select the behavior plan of the vehicle 100 to be tracked, where the behavior may correspond, for example, to driving in the middle of a lane or to a lane change. 1 b shows a highly simplified situation in which vehicle 100 has a specific actual position 135 and actual orientation 140, which are to be converted into a target position 125 and a target orientation 130 along a specified target path 120 as part of the execution of the automatic driving function in vehicle 100. The processing unit 110 forwards the determined desired path parameters 165 or the determined desired path 170, subsequently in the 3 a and 3 b shown, to a control unit 115 further. The control unit 115 can, for example, be an in 1 a include controller structure not shown, which is designed based on the determined target path 170 to regulate the target path parameters relevant to the vehicle guidance and on their basis to control vehicle devices not shown, such as a front wheel steering or rear wheel steering.

2 a FIG. 1 shows a schematic representation of a first specific embodiment of a method 200 for automatically executing a driving function in vehicle 100. FIG 1 a . the 3 a and 3 b represent approximating curves for describing the specified desired path 120. A first method step 205 provides for recording discrete situational data relating to the driving function of the vehicle 100 via the above-mentioned data acquisition unit 105. The discrete situational data, which can be merged with the stationary data and grouped according to an evaluation rule 155, e.g. combined as a polygon, are further processed by the processing unit 110 in a second method step 210, namely in that the target path parameters are determined in two stages, with the specified target path 120 in 3 a is drawn in dashed lines.

können die genannten Sollbahnparameter ermittelt werden, wobei Δ_i,pose(x) = f_pose (p_pase, x) - (m_ix + b_i) ist, d.h. eine Abweichung zwischen einem Segment des Polygonzugs m_ix + b_i und einer approximierenden Kurve f_pose (p_pose, x). Beispielsweise kann hierzu Matrizenalgebra verwendet werden.In a first intermediate step 215 in 2 a determines the processing unit 110 in 1 a from the discrete situation data target path parameters for position and orientation by a first optimization problem, a position, for example, a deviation between an actual position 135 and a target position 125 of the vehicle 100 and an orientation can be angle-dependent, for example a deviation between an actual orientation 140 and a target orientation 130 may include. For example, a vector p _pose is defined for this purpose, which summarizes the target path parameters to be determined for the position and orientation of vehicle 100, ie the states. For example, the p _pose vector can be, but is not limited to, a 3x1 vector. By analytically solving the integral optimization problem $$\underset{p_{pOse}}{min}{\displaystyle \sum _i{\displaystyle \underset{x_i}{\overset{x_{i+1}}{\int }}{\left({\Delta }_{i,pOse}\left(p_{pOse},x\right)\right)}^2\mathrm{i.e}x}}$$

the specified path parameters can be determined, where Δ _i,pose (x)=f _pose ( _ppase , x)-(m _i x +b _i ), ie a deviation between a segment of the polygon m _i x +b _i and an approximating curve f _pose (p _pose , x). For example, matrix algebra can be used for this.

3 a shows the discrete situation data in a Cartesian representation, which has been grouped accordingly by the processing unit 110 according to the evaluation rule 155, which represents, for example, a polygon of the discrete situation data. Alternatively, a Taylor approximation or other combination of the discrete measurement data is also conceivable, for example grouping three measurement points, etc. For example, from the perspective of the moving vehicle 100 in 3 a a starting point of vehicle 100 is selected at x=0 in the vehicle coordinate system, with the axes each including distance information in meters. However, the representation is not limited to the vehicle coordinate system and could also have been based on another reference coordinate system. In order to determine the target path parameters 145 for the position and orientation of the vehicle 100 (or the approximating curve 150 of the target path parameters 145 for the position and orientation), in 3 a the first observation horizon 160 is taken into account. In this case, the approximating curve 150 of the desired path parameters can be obtained from the above relation f _pose (p _pase, x). The first observation horizon 160 corresponds, for example, to the close range, i.e. the local environment of the vehicle 100.

wobei ein zu ermittelnder Vektor p die Sollbahnparameter für den Krümmungsverlauf 165 zusammenfasst und den Vektor p_pose zumindest teilweise umfasst und dabei eine Abweichung zwischen einem Segment des Polygonzugs m_ix + b_i der diskreten Messdaten und einer approximierenden Kurve f(p,x) Δ_i(x) = f(p,x) - (m_ix + b_i) entspricht. Die Sollbahnparameter für den Krümmungsverlauf 165 sowie die approximierende Kurve 170 für die Sollbahnparameter des Krümmungsverlaufs sind in 3 a ebenfalls eingezeichnet. Hierzu wurde beispielsweise ein zweiter Beobachtungshorizont 175 unabhängig vom ersten Beobachtungshorizont 160 gewählt, der sich über den gesamten positiven x-Bereich, zum Beispiel bis x=12, in 3 a erstreckt und einem Fernbereich entspricht. Zum Beispiel erstreckt sich der ersten Beobachtungshorizont 160 bis x=7. Die approximierende Kurve 171 für den Krümmungsverlauf kann nach der Relation f(p,x) erhalten werden.In a second intermediate step 220 in 2 a determines the processing unit 110 from the discrete situation data with a second Optimization problem a curvature of the target path 120 taking into account the target path parameters 145 for position and orientation as a boundary condition, according to the following formula $$\underset{p}{min}{\displaystyle \sum _i{\displaystyle \underset{x_i}{\overset{x_{i+1}}{\int }}{\left({\Delta }_i\left(p,x\right)\right)}^2\mathrm{i.e}x}}$$

wherein a vector p to be determined summarizes the target path parameters for the curvature 165 and at least partially includes the vector p _pose and a deviation between a segment of the polygon m _i x + b _i of the discrete measurement data and an approximating curve f(p,x) Δ _i (x) = f(p,x) - (m _i x + b _i ) corresponds. The target path parameters for the curvature 165 and the approximating curve 170 for the target path parameters of the curvature are in 3 a also marked. For this purpose, for example, a second observation horizon 175 was selected independently of the first observation horizon 160, which extends over the entire positive x-range, for example up to x=12 in 3 a extends and corresponds to a long-distance area. For example, the first observation horizon extends 160 to x=7. The approximating curve 171 for the course of curvature can be obtained using the relation f(p,x).

The approximating curve for the target path parameters of the course of curvature 170 therefore indicates the overall path or the determined target path, which approximately corresponds to the specified target path 120 . The approximating curve 170 agrees with the approximating position and orientation curve 150 in value and derivative at the point x=0. If the description of the curve is linear in the coefficients, we have a quadratic program that can be solved very efficiently using matrix algebra. This is the case, among other things, for polynomials and polynomials in sections (splines). In the present case, for example, a 3rd degree polynomial is used to describe the model, with all terms up to and including the quadratic term being taken into account when solving the first integral optimization problem, for example to improve the accuracy of the model, since the quadratic term contains the information about the curvature of the includes road. This means that three parameters result from the first integral optimization problem, but only two parameters, for example, are used for further processing (solution of the second integral optimization problem). It can also be used for curves in parametric representation, i.e. curves with an explicit description of the coordinates in several dimensions (plane, space), depending on curve parameters, e.g. arc length.

In a third method step 225 in 2 a Finally, the automatic driving function is carried out by the control unit 115 based on the determined curve of the target path parameters of the curvature profile 170 of the vehicle 100 in accordance with the above explanation. In the 3 a and 4 a and 4 b are a third observation horizon 180 ( 3 a) and a fourth observation horizon 185 ( 4 a) and a fifth observation horizon 190 are shown. In this case, for example, the approximating curve 150 of the target path parameters 145 for position and orientation was not shown. For example, the third observation horizon runs from 180 to x=40, the fourth observation horizon from 185 to x=75 and the fifth observation horizon from 190 to x=120. The third to fifth observation horizons 180, 185, 190 each represent long-distance areas and the determined target path parameters for the curvature profile 165 and the determined curve of the target path parameters for the curvature profile 170 are each determined according to the above dynamic optimization. The number of data points for the scoring rule 155 in the 3 a and 3 b and 4 a and 4 b are exemplary and could have been presented differently. Likewise, the observation horizons could have been chosen differently.

2 B FIG. 1 shows a schematic representation of a second specific embodiment of a method 300 for automatically executing a driving function in a vehicle 100. FIG 1 a . A first method step 305 in 2 B corresponds to the first method step 205 in 2 a , which was described above. In contrast to the method 200 in 2 a , a second method step 310 in 2 B in the first step, in which target path parameters for position and orientation 145 are determined from discrete situation data with the aid of the above first optimization problem, which are combined in the vector p _pose , additionally comprise a regularization. In this case, the regularization can be derived from a previous determination of the target path parameters and/or from further situation data relating to the vehicle odometry, such as a speed of the vehicle, a yaw rate, etc. The structure of the regularization, ie the structure of the regularization term, is subsequently determined using a third method step 315 in 2 B explained. In the second step, the third method step 315 includes the determination of the curvature of the target path, taking into account the target path parameters for position and orientation 145 with an additional regularization term, which corresponds to an estimated vector pest. In 2 B so goes the regularization term that depends on. the estimated vector pest is also included in the calculation the target path parameters for the position and orientation 145 in the first method step 210.

wobei eine Abweichung zwischen dem geschätzen Vektor pest und dem zu optimierenden Vektor p als zusätzlicher Term berücksichtigt wird und λ Gewichtungsfaktoren sind, mit denen der Einfluss der Regularisierung gesteuert werden kann. Insbesondere kann die Parameterschätzung als Kalman-Filter ausgebildet sein. The second integral optimization problem reads accordingly $$\underset{p}{min}{\displaystyle \sum _i{\displaystyle \underset{x_i}{\overset{x_{i+1}}{\int }}{\left({\Delta }_i\left(p,x\right)\right)}^2\mathrm{i.e}x+{\lambda }^T{\left(p-p_{est}\right)}^2}}$$

where a deviation between the estimated vector pest and the vector p to be optimized is taken into account as an additional term and λ are weighting factors with which the influence of the regularization can be controlled. In particular, the parameter estimation can be designed as a Kalman filter.

The nominal path parameters for the course of curvature 165, which are contained in the vector p to be determined, are as in 2 B is shown, in a fourth method step 320 in the determination of the regularization term, ie the estimated vector pest. The estimated vector pest, in turn, is included in the two integral optimization problems as part of the regularization term, so that the target path parameters to be determined in a current time step are always based on the target path parameters determined in a preceding time step. In a fifth method step 325, the target path parameters 165 contained in the vector p are transferred to a downstream controller structure in the control unit 115, which processes them further and in a sixth method step 330 executes the automatic driving function as described above.

The invention has been described in detail by preferred embodiments. Instead of the exemplary embodiments described, further exemplary embodiments are conceivable, which can have further modifications or combinations of the features described. For this reason, the invention is not limited by the disclosed examples, since other variations can be derived therefrom by a person skilled in the art without departing from the scope of protection of the invention.

QUOTES INCLUDED IN DESCRIPTION

This list of the documents cited 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 assumes no liability for any errors or omissions.

Patent Literature Cited

• DE 102018206018 A1 [0004]

Claims (9)

Method (200, 300) for automatically executing a driving function in a vehicle (100), discrete situational data relating to the driving function being recorded (205, 305), the discrete situation data being converted into a time-consistent, smooth target path (170) in that target path parameters (145, 165) are determined in two stages (210, 310, 315), wherein in a first step (215, 310) from the discrete situation data with a first optimization problem target path parameters for the position and Orientation (145) are determined, and in a second step (220, 315) a curvature course of the target path (165, 170) is determined from the discrete situation data with a second optimization problem, taking into account the target path parameters for position and orientation (145), wherein the driving function is automatically executed using the determined desired path (170) (225, 330). procedure after claim 1 , wherein in the first step (215, 310) situation data about a first observation horizon (160) and in the second step (220, 315) situation data about a second observation horizon (175, 180, 185,190) are taken into account, wherein the first observation horizon (160) is shorter than the second observation horizon (175, 180, 185,190). procedure after claim 1 or 2 , where a vector p _pose summarizes the target path parameters for position and orientation (145), where a deviation between a segment of a polygon m _i x + b _i and an approximating curve f _pose (p _pose , x) $${\Delta }_{i,pOse}\left(x\right)=f_{pOse}\left(p_{pOse},x\right)-\left(m_{i}x+b_i\right)$$

is, and where to determine the vector p _pose , the integral optimization problem $$\underset{p_{pOse}}{min}{\displaystyle \sum _i{\displaystyle \underset{x_i}{\overset{x_{i+1}}{\int }}{\left({\Delta }_{i,pOse}\left(p_{pOse},x\right)\right)}^2\mathrm{i.e}x}}$$

is resolved. procedure after claim 3 , where a vector p summarizes the target path parameters for the curvature (165), which contains the vector p _pose , where a deviation between a segment of the polygon m _i x + b _i and an approximating curve f(p,x) $${\Delta }_i\left(x\right)=f\left(p,x\right)-\left(m_{i}x+b_i\right)$$

is, where to determine the vector p, the integral optimization problem $$\underset{p}{min}{\displaystyle \sum _i{\displaystyle \underset{x_i}{\overset{x_{i+1}}{\int }}{\left({\Delta }_i\left(p,x\right)\right)}^2\mathrm{i.e}x}}$$

is resolved. Procedure according to one of Claims 1 until 4 , wherein at least one optimization problem is supplemented by at least one regularization term, which is derived from a previous determination of the target path parameters (145, 165) and/or from other situation data. procedure after claim 5 , where the regularization term is weighted. procedure after claim 6 , where the regularization term is an estimated vector pest, and where a deviation between the estimated vector and the vector to be optimized is taken into account as an additional term in the integral optimization problem in the second step with $$\underset{p}{min}{\displaystyle \sum _i{\displaystyle \underset{x_i}{\overset{x_{i+1}}{\int }}{\left({\Delta }_i\left(p,x\right)\right)}^2\mathrm{i.e}x+{\lambda }^T{\left(p-p_{est}\right)}^2}}$$

where λ are the weighting factors. Procedure according to one of Claims 1 until 7 , wherein the automatic driving function is a lateral guidance of the vehicle (100). Vehicle (100) with an automatic driving function with a method according to one of Claims 1 until 8th is controlled.

Images