DE102014223000B4 - Adjustable trajectory planning and collision avoidance - Google Patents

Abstract

Method (700) for determining a trajectory (112) for the longitudinal and/or lateral guidance of a vehicle (100), the method (700) comprising determining (701) values of one or more trajectory parameters (604) , wherein the one or more trajectory parameters (604) indicate a dynamic and/or a target accuracy of the trajectory (112) to be determined; - determining (702) values of one or more weighting factors (502, 503) from the values of the one or a plurality of trajectory parameters (604) by means of a predetermined characteristic function (600), the predetermined characteristic function (600) indicating corresponding values of the one or more weighting factors (502, 503) for values of the one or more trajectory parameters (604); and - determining (703) the trajectory (112) based on a cost function, the cost function depending on the one or more weighting factors (502, 503), wherein • the cost function comprises a dynamic component which depends on an end point in time of the trajectory (112). depends, and the one or more weighting factors (502, 503) include a dynamics weighting factor (503); and/or• the cost function comprises an accuracy component which depends on a deviation of a target point of the trajectory (112) from a target target, and the one or more weighting factors (502, 503) comprise an accuracy weighting factor (503).

Description

The invention relates to a method and a corresponding device for planning a trajectory for a vehicle.

The realization of partially or highly automated driving of a vehicle (particularly a road motor vehicle) is being promoted both in research and increasingly in the automotive industry. A central aspect of partially or highly automated driving is the planning of the vehicle's trajectory that is as optimal as possible, which avoids collisions with other road users.

A variety of driving assistance functions are known that affect the lateral guidance and/or the longitudinal guidance of the vehicle. For example, parking assistance systems with at least automatic lateral guidance are known, in which the steering is taken over by the vehicle without the driver giving steering instructions; Optionally, longitudinal guidance can also be carried out automatically by the vehicle. With a lane keeping assistant, the driver is supported by lateral guidance to keep the vehicle in the recognized lane. With a side collision warning system, the driver is warned by steering wheel vibrations or a steering impulse when a critical approach to objects occurs; the system can also actively support the driver's evasive steering movements.

The different driving assistance functions typically have different requirements for a trajectory that is to be implemented by the vehicle. In particular, the dynamics of a trajectory to be implemented and/or the accuracy of a goal to be achieved with the trajectory can depend on the driving assistance function. In addition, different driving assistance functions may have different requirements in relation to collision avoidance with an obstacle or object.

For example, describes the DE 10 2012 215 562 A1 a method for determining an optimized avoidance trajectory through lane change and/or avoidance support of a motor vehicle. The optimized avoidance trajectory is determined by optimizing a lateral dynamic quality measure, for which a lateral acceleration and/or a lateral jerk of the motor vehicle is or are used.

This document deals with the technical task of providing a method that makes it possible to efficiently take into account different requirements of driving assistance functions on a vehicle trajectory.

The task is solved by the independent claims. Advantageous embodiments include: described in the dependent claims.

According to one aspect, a method for determining a trajectory for the longitudinal and/or lateral guidance of a vehicle is described. The determined trajectory can be used to provide a driving assistance function that intervenes in the longitudinal and/or lateral guidance of the vehicle. The vehicle can be a road vehicle, e.g. B. include a passenger car, a truck or a motorcycle.

The method includes determining values of one or more trajectory parameters, wherein the one or more trajectory parameters indicate a dynamic and/or a target accuracy of the trajectory to be determined. For example, the one or more trajectory parameters may include a dynamics parameter that indicates a target dynamics of the trajectory. In particular, the dynamics parameter can indicate a target time at which a target point of the trajectory is reached. Alternatively or additionally, the one or more trajectory parameters may include an accuracy parameter that indicates a target accuracy with which a target point of the trajectory corresponds to a target target. In particular, the target accuracy can be used to display a maximum permissible relative deviation of the target point of the trajectory from the target target of the trajectory.

The vehicle can include a variety of driving assistance functions, each of which intervenes in the longitudinal and/or lateral guidance of the vehicle. The values of the one or more trajectory parameters can depend on the driving assistance function for which a trajectory is determined. This means you can use of the one or more trajectory parameters, targets of the different driving assistance functions are effectively transferred to a trajectory planning.

The method further includes determining values of one or more weighting factors for a cost function from the values of the one or more trajectory parameters using a predetermined characteristic function. The predetermined identification function can also include a sequence of a plurality of identification functions. The characteristic function or the sequence of characteristic functions can be used to translate the one or more trajectory parameters (and their meaning within a driving assistance function) into one or more weighting factors (and their meaning for a cost function for determining a trajectory). This makes it possible, as part of the driving assistance function, to set values for one or more trajectory parameters that have a direct significance for the dynamics and/or for the target accuracy of the trajectory to be determined, and to determine corresponding values for one or more weighting factors which ensures that during trajectory planning a trajectory is determined which actually has the desired dynamics and/or target accuracy.

The predetermined characteristic function may depend on a variety of test trajectories that were calculated based on the cost function. For example, test trajectories can be determined for different values of one or more weighting factors. Furthermore, corresponding values of the one or more trajectory parameters can be determined for the test trajectories. A characteristic function can then be determined which indicates corresponding values of the one or more weighting factors for values of the one or more trajectory parameters.

The method further includes determining the trajectory using a cost function, the cost function depending on the one or more weighting factors. In particular, a trajectory can be determined which is optimal with respect to the cost function. For example, a large number of trajectories and corresponding values of the cost function can be determined. The trajectory can then be selected and, if necessary, used for the longitudinal and/or lateral guidance of the vehicle, which has an optimal (e.g. minimum) value of the cost function.

The cost function may include a dynamics component that depends on an end point in the trajectory. The one or more weighting factors may include a dynamics weighting factor, where the dynamics weighting factor weights the dynamics component of the cost function. In one example, the characteristic function is such that a value of the dynamics parameter can be converted into a corresponding value of the dynamics weighting factor.

Alternatively or additionally, the cost function can include an accuracy component that depends on a deviation of a target point of the trajectory from a target target. The one or more weighting factors may include an accuracy weighting factor that weights the accuracy component. In one example, the characteristic function is such that a value of the precision parameter can be converted into a corresponding value of the precision weighting factor.

The one or more trajectory parameters can, for example, assume values between a minimum value and a maximum value. The values of the one or more trajectory parameters can then correspond to target values between a minimum target value and a maximum target value of the dynamics and/or the target accuracy of the trajectory to be determined. For example, the minimum value of the dynamics parameter can correspond to a maximum target value of the target dynamics of the trajectory. Furthermore, the maximum value of the dynamics parameter can correspond to a minimum target value of the target dynamics of the trajectory. The relationship between dynamic parameters and target dynamics can be proportional or inversely proportional. The target dynamics can depend on a target end time of a trajectory or correspond to the target end time of a trajectory.

The method may further include determining whether the trajectory collides with an object in an environment of the vehicle. If it is determined that the trajectory collides with the object, a time period until the collision with the object can be determined. The time period can be compared to a threshold value. The threshold value (also referred to as TTC, time-to-collision, threshold value) can depend on the driving assistance function for which a trajectory is determined. The trajectory can then be used for longitudinal and/or lateral guidance of the vehicle if the time period is greater than or equal to the threshold value. A driving assistance function can therefore efficiently define a time for a change in a currently used trajectory. Alternatively or In addition, a change in the trajectory currently used can be effected through superimposed planning.

According to a further aspect, a method for determining a trajectory for the longitudinal and/or lateral guidance of a vehicle is described. The trajectory planning features described in this document can also be used for this method. The method includes detecting an object in an environment of the vehicle based on environmental data relating to an environment of the vehicle. Furthermore, the method includes determining a plurality of trajectories based on a cost function, wherein the trajectories of the plurality of trajectories differ in relation to a target point and/or in relation to a dynamic (e.g. in relation to a target point in time).

The method further includes selecting a first trajectory from the plurality of trajectories depending on a first value of the cost function for the first trajectory. In particular, the trajectory can be selected that has an optimal value of the cost function. In other words, a corresponding plurality of values of the cost function can be calculated for the plurality of trajectories, where the first value corresponds to the value from the plurality of values for the first trajectory from the plurality of trajectories. The first value may be the smallest value of the plurality of values, and thus the first trajectory may have a minimum cost compared to the other trajectories of the plurality of trajectories. Furthermore, it can be determined whether the first trajectory collides with the detected object. If this is determined, then a time period until the collision with the detected object can be determined. The determined period of time can be compared with a threshold value. The first trajectory can then, depending on the comparison, be used for the longitudinal and/or lateral guidance of the vehicle.

This makes it possible to use a trajectory for the longitudinal and/or lateral guidance of the vehicle that includes a collision with another object if this collision is sufficiently distant in the future. This makes it possible to determine trajectories which are more optimal for a specific driving assistance function. The time of deviation from the optimal trajectory can be determined efficiently by the driving assistance function by setting a threshold value. The method therefore makes it possible to parameterize the collision check when determining a trajectory.

The method may further include determining an alternative trajectory if the time until the collision with the detected object is less than the threshold value. The alternative trajectory can have a value of the cost function that is greater than the first value. In other words, a next best alternative trajectory can be determined or selected.

The first trajectory can be selected, for example, at a first point in time. The method may further include determining, at a second time, whether the first trajectory continues to collide with the detected object. The second point in time follows the first point in time. The second point in time can be such that the remaining time until the originally predicted point in time of the collision of the vehicle with the detected object is greater than or equal to the threshold value. In other words, the second point in time can be selected in such a way that the detected object can also evade at the second point in time by changing the trajectory. On the other hand, checking for a collision at the second time point may allow the originally selected first trajectory to be retained. For example, the detected object may have moved between the first time and the second time in such a way that there is no longer a collision with the detected object on the first trajectory. In this case, the first trajectory can be followed further.

On the other hand, the method can further include determining an alternative trajectory if the first trajectory continues to collide with the detected object at the second point in time. It can also be taken into account whether the remaining time from the second point in time until the collision with the detected object is less than the threshold value. The alternative trajectory can be determined in particular (e.g. selected from the large number of trajectories) if the remaining time period is the same as or possibly smaller than the threshold value.

According to a further aspect, a control unit and/or a trajectory planner for a vehicle are described, wherein the control unit and/or the trajectory planner are set up to carry out one of the methods described in this document.

According to a further aspect, a vehicle (eg a passenger car, a truck or a motorcycle) is described which includes a control unit described in this document and/or a trajectory planner described in this document.

According to a further aspect, a software (SW) program is described. The SW program can be set up to run on a processor (e.g. on a control unit) and thereby carry out one of the procedures described in this document.

According to a further aspect, a storage medium is described. The storage medium may include a SW program configured to be executed on a processor and thereby carry out one of the methods described in this document.

It should be noted that the methods, devices and systems described in this document can be used both alone and in combination with other methods, devices and systems described in this document. Furthermore, any aspects of the methods, devices and systems described in this document can be combined with one another in a variety of ways. In particular, the features of the claims can be combined with one another in a variety of ways.

• 1 eine beispielhafte Fahrsituation mit beispielhaften Trajektorien;
• 2 ein Blockdiagram eines beispielhaften Systems zur Quer- und/oder Längsführung eines Fahrzeugs;
• 3 ein Ablaufdiagramm eines beispielhaften Verfahrens zur Ermittlung einer Trajektorie;
• 4 beispielhafte Koordinaten zur Entkrümmung einer Fahrsituation;
• 5a einen beispielhaften Zusammenhang zwischen den Gewichtungsfaktoren einer Kostenfunktion und einer geplanten Zielquerablage;
• 5b einen beispielhaften Zusammenhang zwischen den Gewichtungsfaktoren einer Kostenfunktion und einem geplanten Endzeitpunkt;
• 6 einen beispielhaften Zusammenhang zwischen einem Gewichtungsfaktor einer Kostenfunktion und einem Dynamik-Parameter; und
• 7 ein Ablaufdiagramm eines beispielhaften Verfahrens zur Ermittlung einer Trajektorie.

The invention is further described in more detail using exemplary embodiments. Show it

• 1 an exemplary driving situation with exemplary trajectories;
• 2 a block diagram of an exemplary system for lateral and/or longitudinal guidance of a vehicle;
• 3 a flowchart of an exemplary method for determining a trajectory;
• 4 exemplary coordinates for decurvature of a driving situation;
• 5a an exemplary relationship between the weighting factors of a cost function and a planned target transverse offset;
• 5b an exemplary relationship between the weighting factors of a cost function and a planned end point;
• 6 an exemplary relationship between a weighting factor of a cost function and a dynamic parameter; and
• 7 a flowchart of an exemplary method for determining a trajectory.

As explained at the beginning, the present document deals with the technical task of adapting the planning of a trajectory for a vehicle in an efficient manner to the different requirements of different driving assistance functions.

In this context shows 1 an exemplary traffic situation. The ego vehicle 100 drives on a multi-lane road 101. A vehicle 102 in the same lane of the road 101 in front of the ego vehicle 102 brakes, so that the ego vehicle 100 has to react in order to avoid a collision. The ego vehicle 100 now has the option of staying in the same lane and braking or changing lanes along a trajectory 111 in order to avoid the braking vehicle 102. A large number of different trajectories 111, 112 are possible for the different options. 1 shows three different trajectories 111 for different driving options (avoidance to the left, braking on the same road, reaching to the right). The different trajectories 111 differ, among other things, with regard to the driving dynamics and with regard to the risk of collision with another vehicle 102. Furthermore, it shows 1 different trajectories 112 for a specific driving option. The different trajectories 112 differ in particular with regard to the driving dynamics and the target points of the trajectories 112.

As already explained, a vehicle 100 often has different driving assistance functions that affect the lateral guidance and/or the longitudinal guidance of the vehicle 100. This is in 2 illustrated. 2 shows a system 200 for controlling the longitudinal/lateral guidance of a vehicle 100. The system 200 receives requests for automatic longitudinal/lateral guidance by a driving assistance function 201 from a variety of driving assistance functions 201. In particular, a driving task to be implemented (e.g. a parking maneuver) is sent to a trajectory planning 202 of the system 200. The trajectory plan tion 202 determines a trajectory 111, 112 for implementing the driving task. Specific requirements of the driving assistance function 201 can be taken into account, such as dynamics of the planned trajectory 111, 112, target accuracy of the trajectory 111, 112, and/or requirements for collision avoidance.

The planned trajectory 111, 112 can be described by one or more target trajectory variables (such as a target curvature, a target heading angle, a target transverse offset). In the 2 For example, for lateral guidance, the controller structure shown can cause the target trajectory variables to be controlled by the vehicle 100 (ie by one or more actuators of the vehicle, for example by an electric power steering system 205 (EPS), by a drive and/or by braking of the vehicle 100) can be implemented. For this purpose, the system 200 may include a web guidance controller 203 and a vehicle guidance controller 204. The path guidance controller 203 can be set up to determine a curvature specification or a steering angle specification as a controller output variable for the downstream vehicle guidance controller 204 based on the one or more target trajectory variables and based on one or more corresponding actual trajectory variables. The vehicle control controller 204 can be set up to determine a steering specification for the power steering 205 of the vehicle 100 as a controller output variable based on the curvature specification or the steering angle specification. In an analogous manner, a controller structure can be provided for the longitudinal guidance, which causes the vehicle 100 to implement one or more target trajectory variables (in particular with regard to the driving speed) of a planned trajectory 111, 112.

The planning of a suitable trajectory 112 by the trajectory planner 202 as a basis for the control therefore represents an important aspect for the implementation of the driving task of a driving assistance function 201. 3 shows a flowchart of an exemplary method 300 for determining a trajectory 112 for a vehicle 100. The method 300 can be carried out, for example, by the trajectory planner 202. The determination of a trajectory 112 is preferably carried out in a non-curved coordinate system, relative to the course of the road. The method 300 can therefore include the step of converting state data or values of state variables of the vehicle 100 (such as the position of the vehicle 100, a yaw angle of the vehicle 100 and/or a steering angle of the vehicle 100) from a Cartesian coordinate system into a (decurved) Frenet coordinate system to transform.

The decurvature is exemplary in 4 shown. For the decurvature, measurement signals relating to the state of the vehicle 100 are transformed into a lane coordinate system. The trajectory planning itself does not take place in a Cartesian coordinate system 401, but in a Frenet coordinate system. The Frenet coordinate system is described with respect to a reference curve 400 (eg the middle of the lane of a road). The vehicle position is thus described by the variables s(t) 403 in the longitudinal direction and d(t) 402 in the transverse direction. ṡ(t) and d(t) describe the longitudinal and transverse velocities and s̈(t) and s̈(t) describe the accelerations. Both the vehicle's own movement and the road users to be taken into account are taken into account in the Frenet coordinate system. This transformation clearly corresponds to the decurvature of the coordinate system 401 and thus allows the longitudinal and transverse movement of the vehicle to be optimized separately.

The determination of a trajectory 112 taking into account physical, technical and comfort-oriented boundary conditions results in an optimization problem with secondary conditions. The resulting optimization problem has equation and inequality constraints due to hard boundary conditions (such as driving dynamics limitations, actuator limitations and collision freedom). This fact makes solving the optimization problem a complex problem, especially when the number and shape of the constraints change. This can occur in particular due to the different requirements of different driver assistance functions 201.

To ensure the convergence of the optimization and to determine the global optimum, the convexity of the optimization problem typically has to be proven, which typically cannot be guaranteed due to the existing constraints. The optimization would therefore have to be carried out with different initial conditions in order to ultimately achieve the best result. Furthermore, when using an optimization algorithm, the maximum number of iterations and thus the required computing time is difficult to estimate, which can lead to substantial problems for a control unit of a vehicle 100, which is supposed to calculate trajectories 112 in real time.

For the reasons mentioned, an alternative solution to the optimization problem is used below, which takes into account the structure of the entire optimization problem and the knowledge about it exploits the limitation of possible solutions. This allows a significant reduction in the computational effort to be achieved.

(d.h. die 3^te Ableitung der Zustandsgröße x₁(t)) definiert werden. In diesem Dokument wird jedoch vorgeschlagen, als Eingang die Ableitung des Rucks $x_1^{\left(4\right)}\left(t\right)$

(d.h. die 4^te Ableitung der Zustandsgröße x₁(t)) zu verwenden. Damit können, wie im Folgenden gezeigt, Polynome 7. Ordnung als Ansatz verwendet werden. Insbesondere kann damit die Anzahl der Freiheitsgrade (z.B. die Anzahl der Randbedingungen) erhöht werden, was insbesondere in Bezug auf die Querführung des Fahrzeugs 100 vorteilhaft ist.The transverse and longitudinal movement of a vehicle 100 can be described as an optimal control problem with output s(t) = x ₁ (t) (in the case of longitudinal planning) or d(t) = x ₁ (t) (in the case of transverse planning) of an integrator system (ie a vehicle model). Here x ₁ (t) is a first state variable of the vehicle 100, which describes the position of the vehicle 100 (in the longitudinal direction or in the transverse direction). The jerk can be used as an input to the integrator system $x_1^{\left(3\right)}\left(t\right)$

(ie the ^3rd derivative of the state variable x ₁ (t)). However, in this document it is proposed to use the derivative of the jerk as input $x_1^{\left(4\right)}\left(t\right)$

(ie the ^4th derivative of the state variable x ₁ (t)). This means that 7th order polynomials can be used as an approach, as shown below. In particular, the number of degrees of freedom (eg the number of boundary conditions) can be increased, which is particularly advantageous with regard to the lateral guidance of the vehicle 100.

wobei die Eingangsröße u der Ableitung des Rucks $x_1^{\left(4\right)}\left(t\right)$

entspricht. Der Zustand eines Fahrzeugs 100 zu einem bestimmten Zeitpunkt t kann durch den Zustandsvektor x^T = [x₁, x₂, x₃, x₄] beschrieben werden, wobei x₂(t) = ẋ₁(t), x₃(t) = ẋ₂(t) und x₄(t) = ẋ₃(t).The integrator system used, for example, for cross-planning can be defined as follows: $$\dot{\text{x}}=\left[\begin{array}{cccc}0& 1& 0& 0\\ 0& 0& 1& 0\\ 0& 0& 0& 1\\ 0& 0& 0& 0\end{array}\right]\text{x}+\left[\begin{array}{c}0\\ 0\\ 0\\ 1\end{array}\right]u$$

where the input size u is the derivative of the jerk $x_1^{\left(4\right)}\left(t\right)$

corresponds. The state of a vehicle 100 at a specific time t can be described by the state vector x ^T = [x ₁ , x ₂ , x ₃ , x ₄ ], where x ₂ (t) = ẋ ₁ (t), x ₃ (t ) = ẋ ₂ (t) and x ₄ (t) = ẋ ₃ (t).

It can be shown that the state vector x(t) is given as $$\text{x}\left(t\right)=\underset{=:{\text{M}}_1\left(t\right)}{\underbrace{\left[\begin{array}{cccc}1& t& t^2& t^3\\ 0& 1& 2t& 3t^2\\ 0& 0& 2& 6t\\ 0& 0& 0& 6\end{array}\right]}}{\text{c}}_{0123}+\underset{=:{\text{M}}_2\left(t\right)}{\underbrace{\left[\begin{array}{cccc}{t}^4& t^5& t^6& t^7\\ 4t^3& 5t^4& 6t^5& 7t^6\\ 12t^2& 20t^3& 30t^4& 42t^5\\ 24t& 60t^{2}21& 120t^3& 210t^4\end{array}\right]}}{\text{c}}_{4567}$$

The above equations describe a 7th order polynomial with respect to the spatial course x ₁ (t).

To solve the optimization problem based on a 7th order polynomial, the following quality functional (also referred to as a cost function) can be used: $$J=\frac{1}{2}{\displaystyle \int {}_0^{t_f}}{\left(x^{\left(4\right)}\left(t\right)\right)}^{2}dt+k_1{\left(x_{ref}-x_1\right)}^2+k_2{t}_f$$

The parameters c ₀₁₂₃ ^T = [c ₀ , c ₁ , c ₂ , c ₃ ] are calculated from the initial conditions x(0) = x ₀ of the trajectory 112 at time t = 0 as $${\text{c}}_{0123}={\text{M}}_1^{-1}\left(0\right){\text{x}}_0$$

The parameters c ₄₅₆₇ ^T = [c ₄ , c ₅ , c ₆ , c ₇ ] are calculated from the end conditions x(t _f ) of the trajectory 112 at time t = t _f as $${\text{c}}_{4567}={\text{M}}_2^{-1}\left(t_f\right)\left(\text{x}\left(t_f\right)\right)-{\text{M}}_1\left(t_f\right){\text{c}}_{0123}$$

The end conditions can be specified as in the above formula. Alternatively, a reference course described by x _ref = [x _1,ref , x _2,ref , x _3,ref , x _4,ref ] ^T can be specified. The optimization goal In this case, the aim is to get as close as possible to this reference curve. In this case, the parameters c ₄₅₆₇ ^T can be calculated as follows: $$\begin{array}{l}{\text{c}}_{4567}={\text{M}}_3{\left(t_f\right)}^{-1}\left({\text{x}}_{ref}-{\text{M}}_1\left(t_f\right){\text{c}}_{0123}\right)\\ \text{with}\\ {\text{M}}_3\left(t_f\right)={\text{M}}_2\left(t_f\right)-\left(\begin{array}{cccc}0& 0& 0& \frac{5040}{k_2}\\ 0& 0& 0& 0\\ 0& 0& 0& 0\\ 0& 0& 0& 0\end{array}\right).\end{array}$$

The optimization problem now consists in determining both the end time or the end time t _f and, if necessary, the end state x(t _f ) for an optimal trajectory 112.

In particular, to calculate a transverse trajectory, 7th order polynomials can be selected in order to be able to specify the 3rd derivative of the trajectory 112 at the beginning and end of the trajectory 112. In this way, a steering angle at the beginning and at the end of the trajectory 112 can be specified much more precisely. A target area d _target can be used as the desired end point of a trajectory 112, which, for example, corresponds to a selected driving option (see 1 ) corresponds. This target area can, for example, be defined as the end position of the final state x(t _f ), with x ₁ (t _f ) = d _target . In particular, different trajectories 112 can be determined in a target window, the target window corresponding to a selected driving option.

The following function can be used as a selection measure or as a quality measure for determining a trajectory 112 (also referred to as a cost function) for the lateral guidance of the vehicle 100: $$J_{quer}=\frac{1}{2}{\displaystyle \int {}_0^{t_f}{\left(d^4\left(t\right)\right)}^{2}dt+k_{q1}{\left(d_{\mathrm{e.g}iel}-d\left(t_f\right)\right)}^2+k_{q2}{t}_f}$$

The first expression evaluates the development of the derivative of the jerk along the trajectory 112. The second expression evaluates the deviation of the final position d(t _f ) from the target position d _target and can also be referred to as the accuracy component of the cost function. Furthermore, the third expression evaluates the time length of the trajectory 112 (and thus the dynamics of the trajectory 112) and can also be referred to as the dynamics component of the cost function. The expression of the trajectory 112 can be weighted using the weighting factors k _q1 and k _q2 . The weighting factor k _q1 can be referred to as the precision weighting factor and the weighting factor k _q2 as the dynamics weighting factor.

Longitudinal planning can be done in a similar manner. However, it has been shown that a 4th or 5th order polynomial is sufficient for longitudinal planning. Furthermore, combined longitudinal and transverse planning can be carried out by combining (e.g. summing) the selection dimensions for the transverse guidance and for the longitudinal guidance.

To determine an optimal trajectory, the selection measure (ie the cost function) J can be calculated for different values of t _f (ie for different dynamics) and/or for different end states x(t _f ) (ie for different goals or target accuracies). The above formulas can be used for this. This results in a progression of the selection measure J, whereby a minimum value of the selection measure J indicates the optimal value for the time t _f and/or for the final state x(t _f ). In a further step, one or more additional conditions can then be taken into account.

The additional conditions can therefore be taken into account downstream of the optimization. For this purpose, the best trajectory 112 in terms of the cost function J (i.e. the selection measure) can be selected and checked for compliance with the secondary conditions. If the secondary conditions are met, the trajectory 112 is implemented. Otherwise, the next best trajectory 112 is selected and checked for the constraints. This procedure is used until an optimal solution is found that meets the additional conditions.

Actuator and driving dynamics limitations can be taken into account as secondary conditions. Furthermore, the freedom from collision with predicted object trajectories of other objects/vehicles 102 can be taken into account.

des Fahrzeugs 100. Dabei gibt $x_1^{\left(3\right)}\left(t\right)$

die dritte Ableitung der Position x₁(t) des Fahrzeugs 100 an.This in 3 Illustrated method 300 for determining a trajectory 112 for the longitudinal and/or transverse guidance of a vehicle 100 includes determining 301 initial values or initial conditions x(0) = x ₀ for a plurality of state variables x of the vehicle 100. The plurality of state variables includes For example, a position x ₁ (t) of the vehicle 100, a speed ẋ ₁ (t) of the vehicle 100, an acceleration ẍ ₁ (t) of the vehicle 100 and a jerk $x_1^{\left(3\right)}\left(t\right)$

of the vehicle 100. There is $x_1^{\left(3\right)}\left(t\right)$

the third derivative of the position x ₁ (t) of the vehicle 100.

The method 300 also includes determining 302 final values x(t _f ) at an end time t _f for the plurality of state variables x of the vehicle 100. The method 300 also includes determining 303 a trajectory 112 based on the initial values x(0) = x ₀ , the final values x(t _f ), the final time t _f and based on a polynomial with 7th or higher/lower order. The 7th or higher/lower order polynomial may describe the position x ₁ (t) of the vehicle 100 as a function of the time t between the initial position x ₁ (0) and the final position x ₁ (t _f ) of the trajectory 112. To calculate the (state) trajectory x(t) 112, the formulas given above for c ₀₁₂₃ ^T , c ₄₅₆₇ ^T and x(t) can be used. In particular, the formula for x(t) describes the position x ₁ (t) of the vehicle 100 by a 7th order polynomial.

To determine an optimal (state) trajectory 112 (in the sense of a selection measure J), the values of the selection measure J for (state) trajectories 112 with different end times t _f and/or with different end values x(t _f ) of the multitude of state variables be determined. The (state) trajectory 112 can then be selected, which optimizes the selection measure J. Furthermore, it can be checked whether one or more additional conditions are met (as explained above).

The determined trajectory 112 can finally be transformed from the Frenet coordinate system back into a Cartesian coordinate system. Furthermore, the determined trajectory 112 can be used to guide the vehicle 100 using the system 200 (e.g. for an evasive maneuver or for a parking maneuver).

As explained above, different driving assistance functions 201 have different requirements for the driving dynamics and/or the target accuracy of a trajectory 112 to be driven. These parameters “driving dynamics” and “target accuracy” can in principle be changed using the weighting factors k _q1 and k _q2 . However, it is unclear how changing a weighting factor k _q1 , k _q2 affects the driving dynamics and/or the target accuracy.

5a shows an exemplary relationship between the end position d(t _f ) 501 of a trajectory 112 and the values of the weighting factors k _q1 503 and k _q2 502. It can be from 5a It can be seen that, in particular, by increasing the weighting factor k _q1 503, the end position d(t _f ) can be caused to have increased accuracy in relation to a target position d _target . Furthermore, the implementation of a higher target position d _target 501 can be achieved by increasing the weighting factor k _q1 503.

Shows in an analogous way 5b an exemplary relationship between the end time t _f 504 of a trajectory 112 and the values of the weighting factors k _q1 503 and k _q2 502. Off 5b In particular, it can be seen that the end time t _f 504 to be achieved primarily depends on the weighting factor k _q2 502. In particular, by increasing the weighting factor k _q2 502, earlier or shorter end times t _f 504 can be brought about. In other words, an increase in the dynamics of a trajectory 112 can be achieved by increasing the weighting factor k _q2 502.

6 shows a characteristic curve (also referred to as a characteristic function) 600 through which a dynamic parameter 604 can be converted into a weighting factor k _q2 502 (or vice versa). The relationship between dynamic parameter 604 and weighting factor k _q2 502 can also be described by a sequence of characteristic functions, for example by a first characteristic function, to convert a dynamic parameter 604 into a desired target value (for example into the desired end time t _f 504). to convert, and into a second characteristic function in order to convert the desired target size into a weighting factor k _q2 502, through which an optimal trajectory 112 can be determined, which has (at least approximately) the desired target size.

The characteristic curve 600 (or the second characteristic function mentioned above) can be based on the measured relationships 5a and/or 5b can be determined. The dynamics parameter 604 can be standardized to values between 0 and 1. For example, a fixed value can be selected for a target position d _target (in particular for a transverse shelf), so that d _target /t _f results in a measure for the dynamics of a trajectory 112 (in particular for the transverse dynamics of a trajectory 112). The dynamics parameter 604 can, for example, be proportional to d _target /t _f , and thus display an average dynamics of a trajectory 112. The dynamics parameter 604 with a value of 0 can indicate a minimum average dynamics and a value of 1 can indicate a maximum average dynamics. Between 0 and 1, the mean dynamics can increase linearly from the minimum mean dynamics to the maximum mean dynamics.

A driving assistance function 201 can thus specify a value for the dynamics parameter 604 in order to specify an average dynamics of a trajectory 112. Using the characteristic curve 600, a corresponding value of the weighting factor k _q2 502 can then be determined and used in determining a trajectory 112 by the method 300. This makes it possible to ensure that the resulting trajectory 112 has the desired average dynamics.

In an analogous manner, a characteristic curve can be provided that defines a target accuracy in relation to a target position d _target . The accuracy parameter can be standardized between 0 and 1. The characteristic curve can, for example, be based on the measured relationships 5a and/or 5b can be determined. The characteristic curve can provide a corresponding weighting factor k _q1 503 for an accuracy parameter. The accuracy parameter can indicate (relative) deviations in a linearly increasing manner between a minimum (relative) deviation and a maximum (relative) deviation. For example, the value 0 of the accuracy parameter may indicate the maximum (relative) deviation and the value 1 of the precision parameter may indicate the minimum (relative) deviation.

A driving assistance function 201 can then specify an accuracy parameter (also referred to as a target accuracy parameter), which is converted into a weighting factor k _q1 503 using a characteristic curve. The weighting factor k _q1 503 can then be used in method 300 to determine a trajectory 112 that has the desired target accuracy (ie, deviation).

Alternatively or additionally, a combined characteristic curve can be provided (based on the measured relationships 5a and/or 5b), through which values for the weighting factors k _q1 503 and k _q2 502 are provided for the values of a dynamic parameter and a target accuracy parameter. These values can then be used in the cost function J _across to determine a trajectory 112. The dynamics parameter and the targeting accuracy parameter can also be collectively referred to as trajectory parameters.

As already explained, a transparent application of the leadership behavior of a vehicle guidance function or driving assistance function 201 to be developed is an important requirement for the control structure. A dynamic parameter enables a linear application of the dynamics. The disruptive behavior of the system 200 is typically only marginally influenced and, in return, the application of the disruptive behavior typically only marginally influences the management behavior. This allows for transparent and separately adjustable management dynamics of the System 200.

The guidance behavior is largely determined by the planned trajectory 112, the dynamics of which are determined by the weighting factors k _q1 503 and k _q2 502 of the quality functional J _transverse for the transverse guidance, and by corresponding weighting factors of a quality functional for the longitudinal guidance. An example is given in the 5a , 5b and 6 the application of the lateral guidance trajectory is considered. The procedure can be transferred in an analogous manner to the application of longitudinal guidance trajectories. 5a and 5b show the effect of changing the weighting factors k _q1 503 and k _q2 502 (at intervals [0, 200]) on the achieved target transverse offset 501 and on the time t _f required for this. The target transverse position was chosen as d _target = 3m ± 0.5m and for the permitted time t _f ∈ [0.4, 6] sec. The 5a and 5b A variety of determined test trajectories 112 can be provided based on this.

The 5a and 5b show that the time required to dismantle the transverse shelf is determined in particular by the weighting factor k _q2 502. The weighting factor k _q1 503 primarily influences the achieved target transverse offset 501 in accordance with the quality functional. The target transverse offset 501 is also relative strongly dependent on the selected weighting factor k _q2 502. As in 5a shown, in order to achieve the target transverse offset 501 it is necessary to choose the weighting factor k _q1 503 sufficiently high. This means that the desired dynamics can then be set largely independently using the weighting factor k _q2 502.

5b and 5 show that the relationship between the time 504 to reach the final transverse shelf 501 and the weighting factor k _q2 502 corresponds approximately to the behavior of 1/k _q2 . This behavior is largely independent of the weighting factor k _q1 503. This fact allows a linear relationship between a target time 504 and an abstract dynamic parameter 604 to be achieved. For this purpose, the mean value of the group is taken from 5b as a basis and the resulting map is inverted.

By using characteristic curves 600 and an abstract dynamic parameter and/or an abstract accuracy parameter, a transparent application for different driving assistance functions 201 is achieved. The methodology described can also be transferred to longitudinal guidance. For longitudinal guidance, the two weighting factors of a cost function for longitudinal guidance during free travel and for longitudinal guidance during subsequent travel are also largely independent.

7 shows a flowchart of an exemplary method 700 for determining a trajectory 112 for the longitudinal and/or lateral guidance of a vehicle 100. The method 700 includes determining 701 values of one or more trajectory parameters 604, wherein the one or more trajectory parameters 604 display a dynamic and/or a target accuracy of the trajectory 112 to be determined. The one or more trajectory parameters 605 can in particular include a dynamics parameter and/or a targeting accuracy parameter. The method 700 further includes determining 702 values of one or more weighting factors 502, 503 from the values of the one or more trajectory parameters 604 using a predetermined characteristic function or a predetermined characteristic map 600. The method 700 also includes determining, in particular selecting , 703 of the trajectory 112 based on a cost function (also referred to as a quality functional), the cost function depending on the one or more weighting factors 502, 503.

After determining a large number of trajectories 112 using a quality functional, it can be checked which of the large number of trajectories 112 fulfills the one or more secondary conditions for a driving maneuver. In particular, it can be checked whether at least one trajectory 112 from the plurality of trajectories 112 meet the requirements of the driving assistance function 201 with regard to collision avoidance.

With regard to collision avoidance, it is also desirable to provide transparent adjustability of the reaction time to an impending collision. In particular, a too early reaction to an obstacle 102 is typically not understandable for a driver of the vehicle 100. Furthermore, an impending collision can resolve itself again due to the movement of the obstacle 102 (e.g. due to a lane change, a brief displacement of a vehicle in front, etc.).

As part of trajectory planning, one or more trajectories are typically determined and a trajectory 112 that is collision-free is selected. Each trajectory 112 can be checked for collisions with static and dynamic obstacles 102. If a calculated trajectory 112 collides, the earliest collision time at which the trajectory 112 collides can be calculated. If this point in time is below an adjustable limit, the trajectory 112 is recognized as a “collision cannot be avoided” and is typically rejected. Otherwise, the trajectory 112 can be classified as a “colliding but collision-avoidable trajectory”. Such a trajectory 112 can be selected at a specific time. However, it must be checked at a future point in time whether this trajectory 112 is still permitted. If this is not the case, another trajectory 112 must be selected. This information can be transmitted to the driving assistance function 201.

1. a) die Trajektorie 112 eine Kollision mit einem Hindernis 102 aufweist; und
2. b) zu welchem Zeitpunkt (oder nach welcher Zeitdauer) entlang der Trajektorie 112 es zu einer Kollision mit dem Hindernis 102 kommt. Mit anderen Worten, die TTC der Trajektorie 112 kann ermittelt werden.

The method described above enables a driving assistance function 201 to set a time-to-collision (TTC) threshold, through which a point in time can be determined at which a driving maneuver or the change in trajectory is triggered. In particular, it can be determined for a trajectory 112 whether

1. a) the trajectory 112 has a collision with an obstacle 102; and
2. b) at what point in time (or after what period of time) along the trajectory 112 a collision with the obstacle 102 occurs. In other words, the TTC of trajectory 112 can be determined.

The TTC of the trajectory 112 can then be compared with the TTC threshold value (which can be specified by the driving assistance function 201). If the TTC is below the TTC threshold, the trajectory 112 is excluded or rejected. On the other hand, the trajectory 112 can be taken into account and possibly used for the longitudinal and/or lateral guidance of the vehicle 100 if its TTC is equal to or greater than the TTC threshold.

The vehicle 100 can then be guided according to the selected trajectory 112. At a later time (the trajectory planning can be repeated at a predefined frequency of x calculations per second) a new trajectory 112 can be determined. The TTC of a trajectory 112 can then be compared again with the TTC threshold value. If the obstacle 102 continues to exist, this may result in a trajectory 112 that was previously selected having a TTC that is below the TTC threshold at the later time, so that the trajectory 112 is discarded, and another Trajectory 112 is selected (e.g. a trajectory 112 through which the obstacle 102 is avoided).

It is therefore possible to use an optimal trajectory 112 (according to the quality functional) for as long as possible and only select a less optimal (avoidance) trajectory 112 at an adjustable time. For example, there may be a driving situation in which the vehicle 100 hits an obstacle 102 (eg a front vehicle 102, see 1 ) is closing, and there is also a risk of collision with another object 102 (eg with another vehicle 102 in the right lane, see 1 ) is available. The priority is typically to avoid a collision with the first object 102. After avoiding the first object 102, a suitable maneuver for the second object 102 (eg brake, continue to avoid, etc.) can be determined. The maneuver to avoid the first object 102 can still be selected because the TTC of the trajectory 112 is above a predefined TTC threshold. The definition of a TTC threshold value thus enables the sequential series of driving maneuvers that are as optimal as possible.

Another example is the transition from a free trip to a subsequent trip. If a slow vehicle 102 is traveling in front of the ego vehicle 100, deceleration can be made from a target vehicle speed to the slower speed. For this purpose, a group of trajectories 112 is planned, which consists of a mixture of speed-based and location-based trajectories 112. The speed-based trajectories 112 are, for example, planned to the target vehicle speed and thus lead to a future collision. Nevertheless, a driving assistance function 201 can specify that the ego vehicle 100 travels as far as possible at the target speed and is not braked too early. The timing of the transition from a speed-based trajectory 112 to a location-based (decelerated or evasive) trajectory 112 can be adjusted by the TTC threshold. In particular, via the parameterizable “latest acceptable colliding point in time”, i.e. via the TTC threshold value, a speed-based trajectory 112 can be accepted until a defined delay (e.g. via the dynamics parameter) is still possible.

This document describes a parameter (i.e. in particular the TTC threshold value) with which collision avoidance can be made adjustable. The parameter described can be applied independently and independent of the trajectory planning. The parameter enables an improved and comprehensible reaction to obstacles.

Furthermore, parameters (i.e. the trajectory parameters) were described with which the dynamics of a trajectory in particular can be adjusted in a linear and transparent manner. This enables dynamic specifications to be taken into account efficiently in trajectory planning.

The present invention is not limited to the exemplary embodiments shown. In particular, it should be noted that the description and the figures are only intended to illustrate the principle of the proposed methods, devices and systems.

Claims (11)

Method (700) for determining a trajectory (112) for the longitudinal and/or transverse guidance of a vehicle (100), the method (700) comprising, - Determining (701) values of one or more trajectory parameters (604), the one or more trajectory parameters (604) indicating a dynamic and/or a target accuracy of the trajectory (112) to be determined; - Determining (702) values of one or more weighting factors (502, 503) from the values of the one or more trajectory parameters (604) by means of a predetermined characteristic function (600), the predetermined characteristic function (600) being for values of one or more trajectory parameters (604). indicates values of the one or more weighting factors (502, 503) corresponding to several trajectory parameters (604); and - determining (703) the trajectory (112) based on a cost function, the cost function depending on the one or more weighting factors (502, 503), wherein • the cost function comprises a dynamic component which depends on an end point in time of the trajectory (112). depends, and the one or more weighting factors (502, 503) include a dynamics weighting factor (503); and/or • the cost function comprises an accuracy component which depends on a deviation of a target point of the trajectory (112) from a target target, and the one or more weighting factors (502, 503) comprise an accuracy weighting factor (503). Procedure (700) according to Claim 1 , wherein the predetermined characteristic function (600) depends on a plurality of test trajectories that were calculated based on the cost function. Method (700) according to one of the preceding claims, wherein - the one or more trajectory parameters (604) can assume values between a minimum value and a maximum value; and - the values of the one or more trajectory parameters (604) correspond to target values between a minimum target value and a maximum target value of the dynamics and / or the target accuracy of the trajectory (112) to be determined. Method (700) according to one of the preceding claims, wherein the one or more trajectory parameters (604) - include a dynamics parameter that displays a target dynamics of the trajectory (112); and or - include an accuracy parameter that indicates a target accuracy with which a target point of the trajectory (112) matches a target target. Procedure (700) according to Claim 4 with reference back to Claim 3 , whereby - the minimum value of the dynamics parameter corresponds to a maximum target value of the target dynamics of the trajectory (112); and - the maximum value of the dynamics parameter corresponds to a minimum target value of the target dynamics of the trajectory (112). Method (700) according to one of the preceding claims, wherein the method comprises: - Determine whether the trajectory (112) collides with an object (102); - if it is determined that the trajectory (112) collides with the object (102), determining a time period until the collision with the object (102); - Comparing the time period with a threshold value; and - Using the trajectory (112) for the longitudinal and/or lateral guidance of the vehicle (100) if the time period is greater than or equal to the threshold value. Method (700) according to one of the preceding claims, wherein - the vehicle (100) comprises a plurality of driving assistance functions (201) which intervene in the longitudinal and/or lateral guidance of the vehicle (100); and - The values of the one or more trajectory parameters (604) depend on the driving assistance function (201) for which a trajectory (112) is determined. Method for determining a trajectory (112) for the longitudinal and/or transverse guidance of a vehicle (100), the method comprising - detecting an object (102) in an environment of the vehicle (100), based on environment data relating to an environment of the vehicle (100); - Determining a plurality of trajectories (112) based on a cost function, the plurality of trajectories differing in relation to a target point and/or in relation to a dynamic; - Selecting a first trajectory (112) from the plurality of trajectories (112), depending on a value of the cost function for the first trajectory (112); - Determine whether the first trajectory (112) collides with the detected object (102); - if it is determined that the first trajectory (112) collides with the detected object (102), determining a time period until the collision with the detected object (102); - Comparing the time period with a threshold value; and - using the first trajectory (112) for the longitudinal and/or lateral guidance of the vehicle (100), depending on the comparison. Procedure according to Claim 8 , wherein - the first trajectory (112) is selected at a first point in time; - the method further comprises determining, at a second point in time, whether the first trajectory (112) continues to collide with the detected object (102); - the second point in time follows the first point in time; and - the method further comprises determining an alternative trajectory if the first trajectory (112) continues to collide with the detected object (102). System (200) for transverse and/or longitudinal guidance of a vehicle (100), wherein the system (200) is set up to carry out a method according to one of the preceding claims. Software program which is designed to be executed on a processor and thereby implements a method according to one of the Claims 1 until 9 to carry out.

Images