DE102016218121A1 - Control device for planning an at least partially automatic longitudinal and / or transverse guidance - Google Patents

Abstract

The invention relates to a strategic movement planning of vehicles based on locally optimal driving trajectories. The invention relates in particular to a control device and a method for planning an at least partially automated longitudinal and / or transverse guidance of a motor vehicle (ego vehicle) by means of at least one electronic control unit. By appropriate design (in particular programming) of the control unit according to the present invention are determined from the vehicle environment physically derived and permissible driving options or trajectory hypotheses, the driving options as trajectories considering - the movement model of the own vehicle, - collision relevant obstacles and objects (especially other vehicles) and - physical-physical restrictions (in particular restrictions imposed by the given number of lanes, no passing or other traffic regulations, etc.). Furthermore, according to the invention, an optimal driving option or trajectory hypothesis for the longitudinal and / or transverse movement of the vehicle is determined from these driving options or trajectory hypotheses by minimizing at least one defined quality criterion.

Description

The invention relates to a control device for planning an at least partially automatic longitudinal and / or transverse guidance, but preferably for planning a fully automated driving.

The invention is based on known strategic driving decisions, e.g. the lane choice or avoidance of a stop within the own lane, which are so far taken on the basis of simple, very strong approximating estimates. This has the disadvantage that the collision-free but not very comfortable motion options can be precluded in advance by the abstraction of the driving situation / vehicle environment to simple distance measures, especially in safety-critical situations, which is not given a sufficient basis for strategic strategic planning, especially in critical situations. Furthermore, in known methods, handling sensor noise is insufficient. The globally optimal vehicle movement has so far been approximated only inaccurately.

The following literature was considered:

• [1] F. Borrelli, A. Bemporad, and M. Morari. Predictive control for linear and hybrid systems. 2015 ,
• [2] Fletcher, S. Teller, E. Olson, D. Moore, Y. Kuwata, J. How, J. Leonard, I. Miller, M. Campbell, D. Huttenlocher, et al. The MIT-Cornell collision and why it happened. Journal of Field Robotics, 25 (10): 775-807, 2008 ,
• [3] K. Graichen. Methods of optimization and optimal control. Lecture notes, University of Ulm, 2014 ,
• [4] Benjamin Gutjahr and Moritz Werling. Optimum vehicle transverse guidance by means of linear, time-variant MPC. In Workshop Driver Assistance Systems, 2015 ,
• [5] Y. Koren and J. Borenstein. Potential field methods and their inherent limitations for mobile robot navigation. In Robotics and Automation, 1991. Proceedings., 1991 IEEE International Conference on, pages 1398-1404. IEEE, 1991 ,
• [6] M. Papageorgiou. Optimization: static, dynamic, stochastic methods. Springer, 2012 ,
• [7] Georg Tanzmeister, Martin Friedl, Dirk Wollherr, and Martin Buss. Path planning on grid maps with unknown goal poses. In Conference on Intelligent Transportation Systems, 2013 ,
• [8th] Valerio Turri, Ashwin Carvalho, Hong Kong Tseng, Karl Henrik Johansson, and Francesco Borrelli. Linear model predictive control for lane keeping and obstacle avoidance on low curvature roads. In 2013 16th International IEEE Conference on Intelligent Transportation Systems, pages 378-383. IEEE ,
• [9] M. Werling, S. Kammel, J. Ziegler, and L. Gröll. Optimal trajectories for time-critical street scenarios using discretized terminal manifolds. The International Journal of Robotics Research, 31 (3): 346-359, 2012 ,

The invention has for its object to simplify the planning of an at least semi-automatic longitudinal and / or transverse guidance depending on various objects, in particular obstacles as possible, to generalize and perform in real time.

According to the invention this object is solved by the features of the independent claims, while in the dependent claims preferred embodiments of the invention are given.

The terms trajectory hypotheses, movement hypotheses, driving trajectories and driving options are to be understood as synonymous.

The invention relates to a strategic movement planning of vehicles based on locally optimal driving trajectories.

• – des Bewegungsmodells des eigenen Fahrzeugs,
• – kollisionsrelevanter Hindernisse und Objekte (insbesondere anderer Fahrzeuge) sowie
• – fahrphysikalischer Beschränkungen (insbesondere Beschränkungen durch die vorgegebene Anzahl von Fahrspuren, Überholverboten oder anderen angezeigten Verkehrsvorschriften usw.)

formuliert werden. Weiterhin wird erfindungsgemäß aus diesen Fahroptionen bzw. Trajektorienhypothesen eine optimale Fahroption bzw. Trajektorienhypothese für die Längs- und/oder Querbewegung des Fahrzeugs durch Minimierung mindestens eines definierten Gütekriteriums bestimmt. The invention relates in particular to a control device and a method for planning an at least partially automated longitudinal and / or transverse guidance of a motor vehicle (ego vehicle) by means of at least one electronic control unit. By appropriate design (in particular programming) of the control unit according to the invention, depending on the situation, all of the vehicle environment physically derived and permissible driving options or trajectory hypotheses determined, the driving options as trajectories under consideration

• The movement model of the own vehicle,
• - collision-relevant obstacles and objects (especially other vehicles) as well as
• - physical restrictions (in particular restrictions on the number of lanes, overtaking bans or other traffic regulations, etc.)

be formulated. Furthermore, according to the invention, an optimum driving option or trajectory hypothesis for the longitudinal and / or transverse movement of the vehicle is determined from these driving options or trajectory hypotheses by minimizing at least one defined quality criterion.

The invention also relates to at least one electronic control unit, in particular with a functional module, which corresponds to the embodiment the method according to the invention is configured or programmed.

The invention is based on the following considerations:
By optimizing the driving trajectory according to the invention for different movement options based on the consideration of a movement model of the own vehicle, collision-relevant obstacles and objects as well as physical constraints, a better statement can be made about strategic driving decisions. This achieves a generic approach to the well-founded determination of strategic movement decisions while at the same time ensuring comfortable and safe vehicle movement.

The methods known from the literature for optimizing driving trajectories overtax the vehicle control devices in terms of their computing power in the short and medium term. In order to meet newly emerging quality requirements, the present invention profits from the known advantages of the literature, the limited, linear-quadratic optimization for computing efficient solution finding profitable. Here, the selected problem formulation based on the application-specific use of slip variables leads to functional further developments, so that in addition to the prioritized collision avoidance aspects of natural driving behavior can be taken into account in a special way. The testing of the converted vehicle longitudinal and transverse guidance in a real test illustrates the great potential of this approach for automated driving functions. In order to improve comfort and safety in road traffic, the degree of automation of vehicles has steadily increased in recent decades. In order to further increase this degree up to the highly and fully automated driving against the background of the computing power that can be economically integrated into the vehicle, the most efficient calculation methods for trajectory planning, which are developed for dynamic and time-critical traffic situations, are the subject of today's research. In addition, aspects of natural driving behavior are becoming more and more in the focus, so that the goal of novel approaches to trajectory planning is not only the most efficient generation of safe driving trajectories, but also those that are traceable and pleasant for the driver in different traffic scenarios.

• 1. Die Dynamische Programmierung eignet sich vor allem für die Lösung nicht-konvexer Optimierungsprobleme. Wegen des Fluches der Dimensionalität ist jedoch nur die Onlineberechnung von Systemen mit geringer Ordnung möglich, was eine aus Komfortgründen notwendige Planung eines stetig-differenzierbaren Lenkwinkel- oder Längsbeschleunigungsverlaufs in Echtzeit mit heutiger in Fahrzeugen integrierbarer Rechenleistung ausschließt.
• 2. Die Indirekte Methode eignet sich zur Lösung lokaler Optimierungsprobleme und wurde unter bestimmten Voraussetzungen bereits erfolgreich angewandt. Aufgrund der nur schwierig zu berücksichtigenden Nebenbedingungen konnte allerdings lediglich eine suboptimale Lösung des Problems unter großem Rechenaufwand bestimmt werden.
• 3. Im Gegensatz dazu lassen sich bei der Direkten Methode Nebenbedingungen einfach integrieren. Da die Anwendung direkter Methoden bei nichtlinearen Systemen wegen der iterativen Lösungsfindung sehr rechenintensiv und das Konvergenzverhalten des Problems stark abhängig von einer bestimmten Startlösung ist, nutzt die vorliegende Arbeit die Vorteile der direkten Methode bei linearquadratischer Problemformulierung für die Trajektorienoptimierung automatisierter Fahrfunktionen gewinnbringend aus und stellt deren praktische Relevanz am Beispiel der Fahrzeuglängs- und -querführung in aussagekräftigen Fahrsituationen im Realversuch dar.

Different approaches for generating trajectories for automated driving functions are found in the literature. Under certain simplifications, simple heuristics were used in combination with path planning strategies for a small number of isolated traffic scenarios. Since these rule-based approaches are only to be integrated into a general concept under difficult conditions, alternative methods such as potential fields of varying severity were applied. These methods use rather simplified replacement models instead of vehicle models, which greatly limits the consideration of physical limits. In contrast, optimization methods with differential constraints for planning maneuvers have proven to be particularly advantageous. These can be divided into three groups according to how the solution is found:

• 1. Dynamic programming is especially useful for solving non-convex optimization problems. Because of the curse of dimensionality, however, only the online calculation of systems with low order is possible, which precludes a need for comfort planning a continuously-differentiable steering angle or longitudinal acceleration curve in real time with today's integrated in vehicles computing power.
• 2. The Indirect method is suitable for solving local optimization problems and has already been successfully applied under certain conditions. Due to the difficult to consider constraints, however, only a sub-optimal solution of the problem could be determined under great computational effort.
• 3. By contrast, direct conditions can be easily integrated in the direct method. Since the application of direct methods in nonlinear systems is very computationally intensive and the convergence behavior of the problem is highly dependent on a particular starting solution, the present work uses the advantages of the direct method with linear quadratic problem formulation for the trajectory optimization of automated driving functions and presents their practical advantages Relevance using the example of vehicle longitudinal and lateral guidance in meaningful driving situations in the real test dar.

In view of this, the invention focuses on the computationally efficient solution of time-invariant, linear quadratic optimal control problems with constraints, while considering application-specific extensions of the problem formulation. Based on this, advantageous problem formulations were derived by the invention both for the application to the vehicle longitudinal motion and for the planning of the vehicle lateral movement, which stabilize the control loop of the vehicle guidance in the sense of a linear, model-predictive control (LMPC) with cyclical solution on a progressing optimization horizon. Due to the special formulation of restrictions, this can result in a natural follow-up behavior and safe stopping from the perspective of the longitudinal guidance be realized taking into account the prevailing traction conditions.

Example application of vehicle longitudinal guidance:
For efficient trajectory optimization of the vehicle longitudinal movement, the invention provides a novel problem formulation of a linear-quadratic optimization problem with restrictions. The formulation of specific position and acceleration restrictions in combination with the use of slip variables makes it possible to achieve an optimal compromise between a safe and a comfortable longitudinal movement of the vehicle. The longitudinal movement can be described with sufficient accuracy using the time-invariant integrator system with a state vector and an associated manipulated variable. It should be noted that despite the necessary temporal discretization of the system dynamics, the selected system order of the fourth degree ensures a progression of the acceleration a which can be continuously differentiated for reasons of comfort. Based on the states of the system dynamics, deviations from desired motion presets can be considered quadratically in the cost functional via matrices with weighting factors.

For the planning of safe longitudinal motion trajectories, restrictions are formulated for the previously defined system outputs, which can be derived from collision-relevant obstacles and the prevailing traction conditions. Depending on the varying adhesion of the tires with the ground of the road, they can be considered as acceleration restrictions in the movement planning according to the time-variant inequality constraints. In addition, the formulation of time-variant position constraints enables the inclusion of collision-relevant obstacles in an equivalent way. The predicted positions of other road users at discrete future times limit the position of the own vehicle at the same time. Consequently, for all future vehicle positions on the optimization horizon, their permissible values can be limited by the output restrictions of the shape, whereby safety distances can be parameterized via additional terms.

In accordance with the above-mentioned use of slip variables for flexibly influencing the trajectory form, advantageous acceleration profiles with sections of constant acceleration phase can be generated by two-stage acceleration restrictions in combination with a weakly weighted slip variable for reasons of comfort. In the case of large speed transitions, the extension of the safety-relevant limitation to an optimal compromise between a jerk-optimal longitudinal movement and a minimum required acceleration can be parameterized. As a result, comfortable deceleration curves of identical braking duration with sections of constant acceleration phases are achieved by multi-level output restrictions with targeted use of the slip variables.

The algorithm was tested for validation in a prototypically modified test vehicle of the BMW 5 Series for various real maneuvers. For the vehicle longitudinal motion, an optimization step size of 500ms was chosen, so that when choosing N = 20, an optimization horizon of 10.0s resulted. Similarly, an optimization horizon of 4.0s was achieved for the vehicle lateral movement with the optimization step size 200ms. To solve the quadratic program, the open-source solver QLD (www.scilab.org) developed by MJD Powell in 1983 was used. All calculations were done on a dSpace Autobox DS1005 with comparable computing power for current ECUs with a calculation time of 9ms.

In the following a possible reaction of the developed longitudinal guidance to a 40km / h slower vehicle is described. In a first time step, the ego vehicle first keeps the desired speed of e.g. 120km / h and adjusts the speed when einscheren a slower vehicle to this, wherein the optimized acceleration curve characterized by a phase of constant deceleration. This is felt to be particularly pleasant. Subsequently, a comfortable speed adjustment is made at the safety distance within the comfort limits to the front vehicle. In the last time step, the front vehicle shears off, for example, whereupon the algorithm accelerates the vehicle with permissible comfort acceleration of, for example, 1.2 m / s 2 up to the desired speed. The remaining minor deviations in the course of acceleration are due to the switching of the automatic transmission.

As another application, an emergency braking of 80km / h on a wet roadway to a stationary obstacle is described. Already for the first time the algorithm plans a braking maneuver near an acceleration limit. By using the slip variables, the algorithm makes a compromise between parameterizable deviations from the destination and the minimum acceleration. By cyclically solving the optimization problem, the algorithm is able to adapt this compromise to the course of a slightly exaggerated acceleration, so that, unlike initially planned, the deceleration can be conveniently reduced even earlier and the destination is achieved with less deviation.

Summary and Outlook

Based on the linear-quadratic optimization, the present invention derives novel problem formulations both for the vehicle longitudinal and for the vehicle transverse guidance, which can be solved extremely efficiently despite limited computing power and at the same time the high demands on comfort and comfort resulting from the application of automated driving Safety. The application-specific use of slippage variables allows, in addition to aspects of natural driving behavior, such as the minimization of jerking and acceleration, to consider safety-relevant restrictions, such as the avoidance of collisions with static and dynamic obstacles while complying with physical limitations.

Although the algorithms used allow only localized trajectory optimization isolated for the two directions of motion, their highly efficient computation motivates use for longitudinal cross-combined vehicle motion planning. Neglecting the nonlinear feedback effects of vehicle transverse to longitudinal motion, the generation of several locally optimal longitudinal trajectory hypotheses with subsequent optimization of the transverse motion offers the possibility of sufficiently approximating the global solution of the nonlinear trajectory planning problem for different motion hypotheses. In view of this, the focus of future research is on the embedding of this approach in a superordinate driving strategy, which determines the global solution of the trajectory planning problem from locally optimized trajectory hypotheses. In this context, the extension of the vehicle lateral motion model to the linear single-track model promises additional potential for further development of the algorithms.

• 1. Die situationsabhängige Bestimmung aller physikalisch zulässigen Bewegungshypothesen, welche aus der Fahrzeugumgebung ableitbar sind.
• 2. Formulierung der Bewegungshypothesen als lokales Trajektorienoptimierungsproblem was beinhaltet: a. Die Berücksichtigung von i. einem Bewegungsmodell des eigenen Fahrzeugs ii. kollisionsrelevanten Hindernissen und Objekten iii. fahrphysikalischen Beschränkungen b. Die isolierte Optimierung der Fahrzeuglängs- und Fahrzeugquerbewegung (optional)
• 3. Bestimmung der strategisch optimalen Trajektorienhypothese sowohl für die Längs- als auch für die Querbewegungstrajektorie durch Minimierung eines kombinierten Gütekriteriums, welches sich aus den kontinuierlichen Kosten der Trajektorienhypothesen und diskreten kontextabhängigen Termen zusammensetzen kann.

The basic idea of the invention is summarized a method for the strategic movement planning of automated vehicles. This process comprises the following process parts:

• 1. The situation-dependent determination of all physically permissible movement hypotheses, which can be derived from the vehicle environment.
• 2. Formulation of the movement hypotheses as a local trajectory optimization problem which includes: a. The consideration of i. a movement model of the own vehicle ii. collision-relevant obstacles and objects iii. physical limitations b. The isolated optimization of vehicle longitudinal and vehicle lateral movement (optional)
• 3. Determination of the strategically optimal trajectory hypothesis for both the longitudinal and the transverse motion trajectory by minimizing a combined quality criterion, which can be composed of the continuous costs of the trajectory hypotheses and discrete context-dependent terms.

Embodiments of the invention are illustrated in the drawings and will be described in more detail below. It shows

1 a schematic representation of an exemplary situation of an ego vehicle with other obstacles that lead based on here three lanes to corresponding trajectory hypotheses or driving options and

2 an exemplary flowchart for the inventive method.

• Zu 1. Verfahrensteil (siehe oben): Bedingt durch die Position anderer Fahrzeuge 2 bis 5 (Boxen mit Pfeil) und Leitplanken (dicke durchgezogene Linien) ergeben sich die in fett und gestrichelt dargestellten physikalisch zulässigen Bewegungshypothesen bzw. Fahroptionen O1 bis O5. Durch weitere Beschränkungen wie Regeln der StVO können bestimmte Hypothesen bzw. Optionen bereits im Vorfeld ausgeschlossen werden, wodurch schon eine erste Reduzierung der Anzahl der Optionen vorgenommen wird. So ist das Rechtsüberholen des Fahrzeugs auf der linken Spur nicht zulässig.
• Zu 2. Verfahrensteil (siehe oben): Die Bestimmung der dargestellten Trajektorienhypothesen O1 bis O5 ergeben sich als Ergebnis eines lokal formulierten Optimierungsproblems isoliert für die Längs- und Querbewegung. Dadurch kann sichergestellt werden, dass jede Hypothese ein Optimum in Bezug auf Komfort und Sicherheit erfüllt, was essentiell für die nachgelagerte Auswahl der abzufahrenden Trajektorie bzw. Fahroption ist. Darüber hinaus lassen sich lokale Optimierungsprobleme mit deutlich reduziertem Rechenaufwand lösen.
• Zu 3. Verfahrensteil (siehe oben): Durch den Vergleich der im Kontinuierlichen optimierten Bewegungskosten aller Trajektorienhypothesen, vorzugsweise in Kombination mit diskreten Entscheidungsgrößen lässt sich generisch und fundiert je nach Gütekriterium genau eine Hypothese bzw. Fahroption auswählen, welche dann abgefahren werden soll.

In 1 becomes the exemplary situation of the ego vehicle 1 on a three-lane highway for the following vehicle environment:

• First procedural part (see above): Due to the position of other vehicles 2 to 5 (boxes with arrow) and crash barriers (thick solid lines), the physically permissible movement hypotheses or driving options O1 to O5 are shown in bold and dashed lines. Further restrictions, such as StVO regulations, can preclude certain hypotheses or options in advance, which means that the first number of options can be reduced. Thus, overtaking the vehicle on the left lane is not permitted.
• Second part of the method (see above): The determination of the illustrated trajectory hypotheses O1 to O5 are the result of a locally formulated optimization problem isolated for the longitudinal and transverse movement. Thereby it can be ensured that each hypothesis fulfills an optimum in terms of comfort and safety, which is essential for the downstream selection of the trajectory or driving option to be traveled. In addition, local optimization problems can be solved with significantly reduced computational effort.
• 3. Part of the procedure (see above): By comparing the continuously optimized movement costs of all trajectory hypotheses, preferably in combination with discrete decision variables, one hypothesis or driving option can be generically and substantively selected depending on the quality criterion, which is then to be traversed.

The invention is particularly advantageous in autonomous or fully automatic driving Apply vehicles, as they always have to make a decision and in case of doubt can not ask for the driver to take over.

Discrete decision items, such as the past number of lane changes made or an upcoming lane end are referred to herein as discrete contextual terms DKT. By way of example, these terms or decision variables are determined on the basis of tunnel detection, sensor quality detection, weather detection or other parameters which spontaneously limit options.

• K1 für die Option O1: Verzögerungskosten und Lenkkosten für eine Verzögerung und für ein Lenken auf die rechte Fahrspur hinter das Objekt bzw. Fahrzeug 2; K2 für die Option O2: Beschleunigungskosten und Lenkkosten für eine Beschleunigung und für ein Lenken auf die linke Fahrspur hinter das Objekt bzw. Fahrzeug 3; K3 für die Option O3: Verzögerungskosten und Lenkkosten bei Verbleiben auf der eigenen Fahrspur hinter dem Objekt bzw. Fahrzeug 4, das möglicherweise von der rechten Fahrspur auf die mittlere (eigene) Fahrspur wechseln könnte, wobei bei gleicher Fahrgeschwindigkeit von Ego-Fahrzeug 1 und Objekt 4 in diesem Fall die Kosten K3 Null und somit am geringsten sein können; K4 für die Option O4: Beschleunigungskosten und Lenkkosten bei einem Überholen und Ausweichen um das Objekt bzw. Fahrzeug 4, das möglicherweise von der rechten Fahrspur auf die eigene Fahrspur wechseln könnte, sowie Verzögerungskosten und Lenkkosten zum Einscheren hinter das Objekt bzw. Fahrzeug 5; K5 für die Option O5: Beschleunigungskosten und Lenkkosten bei einem Überholen des Objekts bzw. Fahrzeugs 2 sowie bei einem Einscheren hinter das Objekt bzw. Fahrzeug 4, das möglicherweise von der rechten Fahrspur auf die eigene Fahrspur wechseln könnte.

The inventive method will be described in more detail below with reference to 2 explains:
In a costing function block 7 the control unit 6 the costs Ki of the individual trajectory hypotheses or driving options O1 to O5 are planned out continuously and in the present exemplary embodiment are, for example, the following:

• K1 for option O1: delay costs and steering costs for a delay and for steering to the right lane behind the object or vehicle 2 ; K2 for option O2: acceleration costs and steering costs for acceleration and for steering the left lane behind the object or vehicle 3 ; K3 for the option O3: Delay costs and steering costs while remaining in own lane behind the object or vehicle 4 , which could possibly switch from the right lane to the middle (own) lane, being at the same driving speed of ego vehicle 1 and object 4 in this case the cost K3 can be zero and thus lowest; K4 for option O4: acceleration costs and steering costs when overtaking and avoiding the object or vehicle 4 which could possibly change from the right lane to its own lane, as well as delay costs and steering costs for shearing behind the object or vehicle 5 ; K5 for option O5: acceleration costs and steering costs when overtaking the object or vehicle 2 as well as when shearing behind the object or vehicle 4 that could possibly switch from the right lane to its own lane.

The costs Ki give thus in particular the energetic expenditure of the ego vehicle 1 with regard to a negative or positive acceleration and / or a steering or with regard to a required longitudinal and / or transverse guidance of each individual local trajectory hypothesis O1 to O5.

In a selection function block 8th the control unit 6 After determining the continuous costs Ki, the trajectory hypothesis or driving option with the minimum cost MIN (Ki), here eg option O3, is selected. Preferably, a discrete context-dependent term DKT is additionally taken into account for selecting the driving option, if available. For example, such a term DKT may be a tunnel or site detection, by which a lane may be omitted as a possible driving option or trajectory hypothesis. For option O3, this term would not be limiting.

In particular, with fully automated driving an option is to be selected in each case. In this case, a discrete context-dependent term DKT can also include an emergency consideration, which in turn can increase the costs Ki and thus - even with increased costs, which may also include the cost of a collision - the option with the lowest cost Ki is selected.

Such an emergency consideration may provide that depends on exceeding defined distance limits of the ego vehicle 1 to an object a corresponding increase in the cost Ki is made by an increased delay required.

QUOTES INCLUDE IN THE DESCRIPTION

This list of the documents listed by the applicant has been 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.

Cited non-patent literature

• F. Borrelli, A. Bemporad, and M. Morari. Predictive control for linear and hybrid systems. 2015 [0003]
• Fletcher, S. Teller, E. Olson, D. Moore, Y. Kuwata, J. How, J. Leonard, I. Miller, M. Campbell, D. Huttenlocher, et al. The MIT-Cornell collision and why it happened. Journal of Field Robotics, 25 (10): 775-807, 2008 [0003]
• K. Graichen. Methods of optimization and optimal control. Lecture notes, University of Ulm, 2014 [0003]
• Benjamin Gutjahr and Moritz Werling. Optimum vehicle transverse guidance by means of linear, time-variant MPC. In Workshop Driver Assistance Systems, 2015 [0003]
• Y. Koren and J. Borenstein. Potential field methods and their inherent limitations for mobile robot navigation. In Robotics and Automation, 1991. Proceedings., 1991 IEEE International Conference on, pages 1398-1404. IEEE, 1991 [0003]
• M. Papageorgiou. Optimization: static, dynamic, stochastic methods. Springer, 2012 [0003]
• Georg Tanzmeister, Martin Friedl, Dirk Wollherr, and Martin Buss. Path planning on grid maps with unknown goal poses. In Conference on Intelligent Transportation Systems, 2013 [0003]
• Valerio Turri, Ashwin Carvalho, Hong Kong Tseng, Karl Henrik Johansson, and Francesco Borrelli. Linear model predictive control for lane keeping and obstacle avoidance on low curvature roads. In 2013 16th International IEEE Conference on Intelligent Transportation Systems, pages 378-383. IEEE [0003]
• M. Werling, S. Kammel, J. Ziegler, and L. Gröll. Optimal trajectories for time-critical street scenarios using discretized terminal manifolds. The International Journal of Robotics Research, 31 (3): 346-359, 2012 [0003]
• DS1005 [0017]

Claims (6)

Method for planning an at least partially automated longitudinal and / or transverse guidance of a motor vehicle (ego vehicle 1 ) by means of at least one electronic control unit ( 6 ), whereby • depending on the situation, all driving options (O1 to O5) which are physically derivable and permissible from the vehicle environment are determined, the driving options (O1 to O5) being trajectories taking into account - the movement model of the own vehicle ( 1 ), - collision-relevant obstacles and objects ( 2 . 3 . 4 . 5 ) and - driving-physical restrictions are formulated, and • from these driving options (O1 to O5) an optimal driving option (O3) for the longitudinal and / or transverse movement of the vehicle (O3) 1 ) is determined by minimizing at least one defined quality criterion (Ki; Ki, DKT). A method according to claim 1, characterized in that the costs (Ki) of the individual driving options (O1 to O5) are specified as a quality criterion. Method according to one of the preceding claims, characterized in that the quality criterion is specified as a combined quality criterion, which is composed of the costs (Ki) of the driving options (O1 to O5) and defined discrete context-dependent terms (DKT). Method according to one of the preceding claims, characterized in that in the selection of the driving options (O1 to O5) and the isolated optimization of the vehicle longitudinal and vehicle lateral movement is taken into account. Control device with an electronic control unit ( 6 ) for a motor vehicle ( 1 ) for planning an at least partially automated longitudinal and / or transverse guidance of the motor vehicle (ego vehicle 1 ), the control unit ( 6 ) such that • depending on the situation, all physically permissible driving options (O1 to O5) can be determined, which can be derived from the vehicle environment, wherein the driving options (O1 to O5) as trajectories taking into account - the movement model of the own vehicle ( 1 ), - collision-relevant obstacles and objects ( 2 . 3 . 4 . 5 ) as well as - physico-physical restrictions can be formulated, and • from these driving options (O1 to O5) an optimal driving option (O3) for the longitudinal and / or transverse movement of the vehicle ( 1 ) can be determined by minimizing at least one defined quality criterion (Ki; Ki, DKT). Function module in the form of a computer program product for at least one electronic control unit ( 6 ) of the control device according to one of the preceding claims.

Images