DE102018203058A1 - Collision risk-prediction unit - Google Patents

Abstract

The invention relates to a collision risk prediction unit (112) as a part of a vehicle travel support apparatus (100) of an ego vehicle (H), the collision risk prediction unit (112) including a collision rate determination subunit (124) having a Collision rate (D16), and a likelihood that a vehicle does not stop determining unit (126) which determines a likelihood value that a vehicle is not stopping (D20), the collision risk prediction unit (124) being arranged thereto to determine a collision risk prediction value (D24) by multiplying the collision rate (D16) by the probability value that a vehicle does not stop (D20).

Description

The present invention relates to a collision risk prediction unit as a part of a vehicle travel assist apparatus which is adapted to be mounted in an ego vehicle and to assist a driver of the ego vehicle in avoiding a collision with another vehicle. in performing a turn-off turn maneuver.

A specific situation in which the collision risk prediction unit according to the present invention is particularly helpful is a situation in which the ego vehicle makes a turn maneuver, crossing at least one lane of oncoming traffic. In the following, therefore, the invention will be explained and described with reference to this specific situation in order to facilitate its comprehensibility. It should be noted, however, that this is by no means intended to limit the invention to application to this specific situation.

Turning maneuvers in which at least one lane of oncoming traffic is crossed occur in both types of traffic systems, the right-hand traffic system and the left-hand traffic system. In right-hand traffic systems, for example in Continental Europe and the United States of America, at least one lane of oncoming traffic is crossed during a left turn maneuver, while in left hand traffic systems, for example in Japan and the United Kingdom, at least one lane of oncoming traffic a right-turn maneuver is crossed. For the sake of simplicity, the invention will be described herein with reference to a right-hand traffic system, particularly when referring to the drawings. To obtain analogue situations for a left-hand traffic system, the situations described with respect to the direction of travel of the ego vehicle for a right-hand traffic system may simply be mirrored just prior to starting the left-turn maneuver.

WO 2004/068165 A discloses a vehicle travel assist device which detects the current relative direction (direction of existence), the current relative distance, and the current relative velocity of the oncoming vehicle. If the current relative distance or the current relative speed exceeds a predetermined range, a brake unit is automatically activated (autonomous emergency braking - AEB). Further, the time for crossing is calculated by dividing the relative distance by the relative velocity.

Prior art vehicle ride support devices, such as those which comprise WO 2004/068165 A are known, do not provide optimal collision avoidance, as they tend to react rather late. If the control unit initiates the braking action too late, then it is only effective for low ego vehicle speeds and larger intersections. If the ego vehicle speed is too high, then it is not possible to stop the ego vehicle in time. This is because shortly after the driver starts to turn left, the ego vehicle will already have entered the oncoming vehicle's lane. Further, if the intersection is too small, then it is not possible to stop the ego vehicle in time. This is because small crossings have narrow traces. Therefore, the two vehicles have a short lateral distance from each other.

In view of the above, it is therefore an object of the present invention to provide an improved collision risk prediction unit as part of an improved vehicle ride support apparatus.

According to the present invention, this object is achieved by a collision risk predicting unit as a part of a vehicle driving support apparatus which is adapted to be mounted in an ego vehicle and a driver of the ego vehicle in avoiding a collision with to assist another vehicle when performing a lane crossing maneuver, the collision risk predicting unit including a collision rate determining subunit and a likelihood that a vehicle is not stopping determining subunit, the collision rate determining subunit being configured to have a collision rate determine as the percentage of those vehicle trajectories of the ego vehicle from a predetermined plurality of vehicle trajectories that result in a collision with the other vehicle, the determination subunit for a probability that a vehicle is not stopping, is arranged to determine a probability value that a vehicle does not stop as the probability that the ego vehicle will continue the lane-crossing turn maneuver without stopping, the collision risk predicting unit thereto is arranged to determine a collision risk prediction value by multiplying the collision rate by the probability that a vehicle will not stop.

In contrast to the prior art, the invention analyzes the actual situation of the ego vehicle according to two different aspects.

According to a first aspect, a collision rate determination subunit is configured to determine a collision rate by performing a path prediction for the ego vehicle based on at least one vehicle path. The vehicle path may represent a pre-turn section in which the ego vehicle approaches the intersection, for example, by a straight line, a turn section in which the ego vehicle performs a lane-crossing turn maneuver, for example, by a circle section having a turn represents a predetermined radius, and include a post-turn section in which the ego vehicle has completed the lane-crossing turn maneuver and exits the intersection, for example represented again by a straight line. The at least one ego vehicle path may be traversed according to a plurality of vehicle speed profiles, resulting in a plurality of ego vehicle trajectories, each representing a position of the vehicle as a function of time. For example, the path of the oncoming vehicle may also be represented by a straight line. The collision rate determining subunit then determines those ego vehicle trajectories that result in a collision with the oncoming vehicle and divides the number of these collision trajectories by the total number of trajectories considered, so as to obtain the collision rate.

According to a second aspect, a likelihood that a vehicle is not stopping determines the probability for the ego vehicle to perform the lane-crossing turn maneuver without stopping at the intersection, thereby exposing the ego vehicle to the risk of collision with the oncoming vehicle Vehicle is suspended.

Lastly, the collision risk is determined by multiplying the collision rate and the no-stop probability.

Vehicle data, for example, sensor data about the position and speed of the oncoming vehicle, may be defined in a local vehicle coordinate system, whereas data regarding the intersection may be defined in a global coordinate system. For combining these data, the data defined in the global or crossing-fixed coordinate system may be transferred to the local or vehicle-fixed coordinate system originating at the center of the front of the ego vehicle and having its axes in the direction of travel of the ego vehicle (Y-axis) and orthogonal to extend (X-axis). Since it is time-consuming to transfer the entire movement histories of the ego vehicle and the oncoming vehicle to access thereto into the vehicle-mounted coordinate system valid for a given time, the collision risk determination unit may be configured to set the vehicle-mounted one To "freeze" the coordinate system under certain conditions, that is to transform it into a crossing-proof coordinate system. Upon freezing the coordinate system, the previous motion histories of the ego vehicle and oncoming vehicle of the last freeze coordinate system may be deleted because only the future motion history in the coordinate system after freeze is relevant to determining the collision risk. A freezing of the coordinate system can be repeated until an oncoming vehicle turns out to be relevant to collision.

As already mentioned, the collision rate determination subunit may be configured to determine the collision rate based on a vehicle trajectory, wherein the vehicle trajectory is a predetermined lane crossing lane which is traversed according to a predetermined velocity profile. A speed profile should be understood to refer to changing a speed of the ego vehicle along an associated lane crossing lane. That is, one and the same lane crossing lane may be traversed using different velocity profiles resulting in different times and / or time periods for which the collision risk prediction unit may or may not determine a collision with the other vehicle. For example, a plurality of typical no-stop velocity profiles may have been taken during test runs. The velocity profiles are then stored in a memory unit which may or may be associated with the collision risk prediction unit of this invention. According to the velocity profile, the length and / or the curvature of the Bending portion and / or the transition from the pre-bending portion in the bending portion and / or the transition from the bending portion are modified in the post-bending section.

Since vehicles are not points that move along the respective vehicle path, but have a given dimensional extent, a size of the vehicle must be taken into account when determining the potential collision risk. In order to limit the required computation time, the ego vehicle and oncoming vehicle may be represented by a group of three circles, a first circle representing the front of the respective vehicle, a second circle representing the center of the respective vehicle and a third circle representing the rear of each vehicle. Of course, a vehicle is a three-dimensional object, but since vehicles drive on the same road, that is, at the substantially same height level, the extent of the vehicle in its height direction need not be considered for collision risk prediction.

The diameter of the circles may preferably be the same for all circles. Further, the diameter of the circles may be equal to the vehicle width. The position of the center of each circle may be positioned at a longitudinal center plane of the corresponding vehicle, that is, a plane dividing the width of the vehicle into two equal parts. The position of the center of the first and / or third circle may be such that an area extending across the front of the vehicle may be substantially equal to an area which is not through the circle at the corners of the vehicle is covered. The position of the center of the second circle may be such that it is positioned at a transverse center plane of the corresponding vehicle, that is to say a plane which divides the length of the vehicle into two equal parts.

Alternatively or additionally, the position of the center of the first circle may be positioned at 1/6 of the length of the vehicle, and the position of the center of the third circle may be positioned at 5/6 of the length of the vehicle. The position of the center of the second circle can be positioned at 3/6 of the length of the vehicle.

Advantageously, the ego vehicle and the oncoming vehicle may be additionally represented by a further group of three circles, a first circle representing the front, a second circle representing the center, and a third circle representing the rear of the represents the respective vehicle. The diameter of the further group of circles may be greater than that of the aforementioned group of circles. For example, the diameter of the further group of circles may be the diameter of the group of smaller circles plus a predetermined distance, for example 1.0 meter. The diameter of the further group of circles may also be such that the vehicle is completely covered by the area of the three circles associated with the vehicle. By providing the further group of circles, position inaccuracies, for example due to sensor quality, map data inaccuracies or the like, can be compensated.

The collision rate determination subunit may be further configured to collision between the ego vehicle and the oncoming vehicle based on an overlap of the smaller circle representations for both vehicles, based on an overlap of the larger circle representations for both vehicles and based on an overlap of the to check a smaller circle representation for one of the vehicles with the larger circle representation for the other one of the vehicles. In other words, if there is no overlap of the larger circle representation for both vehicles, it may be determined by the collision rate determining subunit that no collision will happen. If there is no overlap of the smaller circle representation for one of the larger circle representation vehicles for the other of the vehicles, it may be determined by the collision rate determination subunit that a collision is unlikely to occur. If there is an overlap of the smaller circle representation for one of the vehicles with the larger circle representation for the respective other of the vehicles, it may be determined by the collision rate determination subunit that a collision is likely to occur. And, if there is an overlap of the smaller circle representation for both vehicles, it may be determined by the collision rate determination subunit that a collision will take place.

In this connection, it should be noted that in order to determine collisions between a first vehicle and a second vehicle on the basis of an overlap of a specific circle representation of the first vehicle with a specific circle representation of the second vehicle, it is for each circle the specific circle representation of the first vehicle is determined to overlap with any of the circles of the specific circle representation of the second vehicle.

Thus, the collision rate determining subunit may be configured to detect collisions that have been determined based on an overlap of the larger circle representations for both vehicles with a lower weighting factor than collisions based on an overlap of the smaller circle representation for one of the vehicles and on the basis of an overlap of the smaller circle representation for one of the larger circle representation vehicles for the respective other of the vehicles with a lower weighting factor than collisions based on an overlap of the smaller ones Circle representations for both vehicles have been determined to weight. As mentioned above, any one of the three possible overlaps of the circle representations may indicate a different level of collision probability. As a consequence, an overlap of the smaller circle representations for both vehicles with a weighting factor of 1.0 may be considered, an overlap of the smaller circle representation for one of the vehicles with the larger circle representation for the other of the vehicles may be weighted at a factor of 0 8, and an overlap of the larger circle representations for both vehicles may be considered with a weighting factor of 0.6.

wobei gemäß dem oben erwähnten Beispiel Σc_small die Anzahl von Kollisionen/Überlappungen zwischen kleinen Kreisen ist, Σc_{small_large} die Anzahl von Kollisionen/Überlappungen zwischen einem kleinen Kreis und einem großen Kreis ist, Σc_large die Anzahl von Kollisionen/Überlappungen zwischen großen Kreisen ist, K_small der Gewichtungsfaktor für Kollisionen/Überlappungen zwischen kleinen Kreisen ist, zum Beispiel 1,0, K_{small_large} der Gewichtungsfaktor für Kollisionen/Überlappungen zwischen kleinen Kreisen und großen Kreisen ist, zum Beispiel 0,8, und K_large der Gewichtungsfaktor für Kollisionen/Überlappungen zwischen großen Kreisen ist, zum Beispiel 0,6.The weighted collision rate can then be calculated by the following formula: $$KOllisiOnstrate=\frac{K_{small}*{\displaystyle \Sigma c_{small}+K_{small\_larGe}*{\displaystyle \Sigma c_{small\_larGe}+K_{larGe}*{\displaystyle \Sigma c_{larGe}}}}}{\text{Number of all variations of the ego}-\text{vehicle}-\text{trajectories}}$$

where, according to the above-mentioned example, Σc _{small is} the number of collisions / overlaps between small circles, Σc _{small_large is} the number of collisions / overlaps between a small circle and a large circle, Σc _{large is} the number of collisions / overlaps between large circles, K _{small is} the weighting factor for collisions / overlaps between small circles, for example 1.0, K _{small_large is} the weighting factor for collisions / overlaps between small circles and large circles, for example 0.8, and K _{large is} the weighting factor for collisions / overlaps between large circles is, for example, 0.6.

In one embodiment of the present invention, for a likelihood that a vehicle is not stopping, the determination subunit may include a plurality of trained decision tree sub-units each configured to output an individual likelihood value that a vehicle is not stopping, and a voting sub-unit is configured to determine the likelihood that a vehicle will not stop based on the plurality of individual likelihood values that a vehicle is not stopping. The dial sub-unit may be configured to determine the probability value that a vehicle is not stopping as the average value of the plurality of individual probability values that a vehicle is not stopping.

• - eine Geschwindigkeit des Ego-Fahrzeugs;
• - eine Geschwindigkeitsänderung des Ego-Fahrzeugs;
• - eine Beschleunigung des Ego-Fahrzeugs;
• - eine Beschleunigungsänderung des Ego-Fahrzeugs;
• - eine Bremsdruckänderung des Ego-Fahrzeugs;
• - eine Zeit des Ego-Fahrzeugs, um mit der erfassten Verzögerung anzuhalten;
• - eine Distanz zu einer Konfliktzone, das heißt die Zone der Kreuzung, in welcher eine Kollision am wahrscheinlichsten ist, aufzutreten;
• - eine Distanz zu einer Konfliktzone nach einem Anhalten mit einer erfassten Verzögerung;
• - eine Distanz zu einer Konfliktzone nach einem Anhalten mit einem komfortablen Bremsen;
• - einen Lenkwinkel des Ego-Fahrzeugs; und
• - eine Lenkwinkeländerung des Ego-Fahrzeugs.

Each of the trained decision tree sub-units may consider at least one of the following parameters:

• a speed of the ego vehicle;
• a speed change of the ego vehicle;
• an acceleration of the ego vehicle;
• an acceleration change of the ego vehicle;
• a brake pressure change of the ego vehicle;
• a time of the ego vehicle to stop with the detected delay;
• - a distance to a conflict zone, that is, the zone of the intersection in which a collision is most likely to occur;
• a distance to a conflict zone after stopping with a detected delay;
• a distance to a conflict zone after stopping with a comfortable braking;
• a steering angle of the ego vehicle; and
• a steering angle change of the ego vehicle.

Frage:

Ist die Geschwindigkeit des Ego-Fahrzeugs über 6,69 m/s?

Antwort:

Ja

Frage:

War die Beschleunigung des Ego-Fahrzeugs innerhalb der letzten 0,4 Sekunden über 0,1 m/s²?

Antwort:

Ja

Frage:

Ist die Geschwindigkeit des Ego-Fahrzeugs über 82,8% des momentanen Geschwindigkeitslimits?

Antwort:

Ja

Ergebnis:

Die Wahrscheinlichkeit, dass das Fahrzeug nicht anhält, wird bestimmt, 100% zu sein

A possible example of a path through a decision tree subunit could be:

Question:

Is the speed of the ego vehicle above 6,69 m / s?

Answer:

Yes

Question:

Was the acceleration of the ego vehicle within the last 0.4 seconds over 0.1 m / s ² ?

Answer:

Yes

Question:

Is the speed of the ego vehicle above 82.8% of the current speed limit?

Answer:

Yes

Result:

The probability that the vehicle will not stop is determined to be 100%

The collision risk prediction unit may be configured to consider environmental structure information. Such environmental structure information may be, for example, the number of lanes in each direction at the intersection approaching the ego vehicle, the presence or absence of an obstacle, for example, a traffic island around which the ego vehicle must travel, and the presence of or Absence of a traffic light.

For this purpose, the environmental structure information can be provided by a navigation system and / or by an e-horizon system and / or by at least one sensor. In particular, environmental structure information including a distance to an intersection and / or an intersection geometry may be based on map data output from a navigation system or obtained by an environmental monitoring sensor. The term "intersection geometry" is intended to describe herein as the number of lanes in the direction of travel of the ego vehicle and / or the number of lanes in the oncoming direction and / or road angles and / or functional road classes.

In addition, it is proposed that the at least one sensor which is attached to the ego vehicle can be in particular a radar system and / or a laser system, for example a lidar system, and / or a camera system. In addition or as an alternative, information may be obtained by a communication device for communication with at least one sensor mounted outside the vehicle, for example, C2X, and or with a communication device formed in another vehicle.

• 1 ein Blockdiagramm einer Fahrzeug-Fahrt-Unterstützungs-Vorrichtung, welche eine Kollisionsrisiko-Vorhersageeinheit umfasst, gemäß der vorliegenden Erfindung zeigt;
• 2 ein Blockdiagramm einer Kollisionsrisiko-Vorhersageeinheit gemäß der vorliegenden Erfindung zeigt;
• 3 ein Flussdiagramm einer Hauptroutine zeigt, welche durch die Fahrzeug-Fahrt-Unterstützungs-Vorrichtung ausgeführt wird;
• 4 ein Flussdiagramm einer Subroutine zeigt, welche durch die Kollisionsraten-Bestimmungsuntereinheit ausgeführt wird;
• 5 ein Flussdiagramm einer Subroutine zeigt, welche durch eine Fahrzeug-Kein-Stopp-Wahrscheinlichkeits-Bestimmungsuntereinheit ausgeführt wird;
• 6 eine Grafik zeigt, welche eine Mehrzahl von typischen Kein-Stopp-Geschwindigkeitsprofilen darstellen;
• 7 eine schematische Ansicht einer spurkreuzenden Abbiege-Situation zeigt, welche ein potentielles Kollisionsrisiko zwischen dem Ego-Fahrzeug und einem anderen Fahrzeug hervorruft;
• 8 ein Kollisionsrisiko-Bestimmungsverfahren von Fahrzeugen durch Repräsentieren der Fahrzeuge als Kreise zeigt.

The invention will be described in greater detail with reference to a specific embodiment with reference to the accompanying drawings, in which:

• 1 FIG. 12 is a block diagram of a vehicle trip assist apparatus including a collision risk prediction unit according to the present invention; FIG.
• 2 shows a block diagram of a collision risk prediction unit according to the present invention;
• 3 Fig. 10 shows a flowchart of a main routine executed by the vehicle travel support apparatus;
• 4 shows a flowchart of a subroutine executed by the collision rate determination subunit;
• 5 shows a flowchart of a subroutine executed by a vehicle no-stop probability determination subunit;
• 6 Fig. 12 is a graph showing a plurality of typical no-stop velocity profiles;
• 7 Figure 11 shows a schematic view of a lane-crossing turn situation which causes a potential collision risk between the ego vehicle and another vehicle;
• 8th shows a collision risk determination method of vehicles by representing the vehicles as circles.

1 shows a block diagram of a vehicle trip support device 100 which is a collision risk prediction unit 112 according to the present invention.

The Vehicle Ride Support Device 100 includes an environmental monitoring unit 102 , which is adapted to the environment of an ego vehicle H to monitor. For example, the environment monitoring unit 102 at least one environmental monitoring sensor 104 who is on the ego vehicle H attached, for example, a camera, a radar system, a lidar system, a GPS system or the like, and / or at least one communication unit 106 which is adapted to communicate with at least one environmental monitoring sensor (not shown) which is external to the ego vehicle H appropriate, for example, using a C2X communication technology.

The output data generated by the environmental monitoring unit 102 be provided to a traffic situation analysis unit 108 forwarded, which is adapted to the current traffic situation of the ego vehicle H relative to another vehicle O to analyze and to determine a value representing the traffic situation for at least one parameter representing the traffic situation, for example the impact factor and / or the time until a collision with the other vehicle O , In addition, the traffic situation analysis unit 108 with sensor units of the ego vehicle H be connected, which generally with 110 for example, a steering wheel angle detecting unit which determines the steering wheel angle of the ego vehicle H indicates a yaw rate detection unit which the yaw rate of the ego vehicle H indicating, and similar detection units, which further operating parameters of the ego vehicle H Show.

Further, the output data generated by the environmental monitoring unit 102 be provided to the collision risk prediction unit 112 (which will be described in greater detail below), which is arranged to predict whether or not there is a risk of colliding with an oncoming vehicle O and output a corresponding collision risk prediction value. The collision risk prediction unit 112 is also with the sensor units 110 of the ego vehicle H connected. Optionally, the collision risk prediction unit 112 also map data from a map data unit 114 receive.

The collision risk prediction value generated by the collision risk prediction unit 112 is issued to a threshold setting unit 116 which is adapted to determine a threshold value for the parameter (s) representing the traffic situation, which is determined by the traffic situation analysis unit 108 is / are determined.

The value (s) of the parameter (s) representing the traffic situation, which are determined by the traffic situation analysis unit 108 is determined, and the threshold / thresholds, which by the threshold setting unit 116 is / are, are to a control / control unit 118 forwarded, which with a brake unit 120 is operatively connected to the control unit 118 to allow a brake command to the brake unit 120 for an automatic braking of the ego vehicle H spend to collide with the oncoming vehicle O to avoid. Optionally, the control unit 118 be operatively associated with other units of assistance, which generally with 122 for example, a driver warning unit, a seatbelt tightening unit, and the like.

In the following, the collision risk prediction unit becomes 112 in greater detail with respect to 2 to be discribed.

The collision risk prediction unit 112 , what a 2 is shown as a dashed box, includes a collision rate determining subunit 124 and a determination subunit for a probability that a vehicle does not stop, 126 ,

The collision rate determination subunit 124 is set up to have a collision rate D16 (please refer 3 and 4 ) as the percentage of those vehicle trajectories of the ego vehicle H from one predetermined plurality of vehicle trajectories, which in a collision with the other vehicle O result. Further details of the collision rate determination subunit 124 will be discussed below with reference to the flow chart of 4 are given.

The determination subunit for a probability that a vehicle does not stop 126 is set up to provide a probability value that a vehicle does not stop, D20 (please refer 3 and 5 ) as the probability to determine that the ego vehicle H the track-crossing turn maneuver will continue without stopping. Further details of the destination subunit for a probability that a vehicle will not stop 126 will be discussed below with reference to the flow chart of 5 are given.

The collision rate D16 generated by the collision rate determination subunit 124 is determined, and the probability that the vehicle does not stop, D20 which is determined by the determination subunit for a probability that a vehicle will not stop 126 is then determined by a multiplication unit 128 multiplied to a collision risk predictive value D24 to determine (see 3 ).

The collision risk prediction unit 112 is set to predict the specific collision risk prediction value D24 to the threshold setting unit 116 to spend (see 1 ).

With reference now to 3 becomes the process performed by the Vehicle Ride Support Device 100 will be described in more detail.

After the ego vehicle H in step S10 has been started, the process goes to a step S12 in which data is collected by sensors, for example radar data, vehicle data and / or map data.

Then the process moves to step S14 in which a route prediction subroutine is executed. The path prediction subroutine will be discussed in greater detail with reference to FIG 4 will be described below. A collision rate D16 is determined by the route prediction subroutine.

Then the process goes to a step S18 in which a maneuver prediction subroutine is executed. The maneuver prediction subroutine will be discussed in more detail below with reference to FIG 5 to be discribed. A no-stop probability D20 is determined by the maneuver prediction subroutine.

Based on the collision rate D16 generated by the route prediction subroutine S14 has been determined, and the no-stop probability D20 generated by the maneuver prediction subroutine S18 has been determined, the risk of collision D24 in one step S22 by multiplying the collision rate D16 and the no-stop probability D20 calculated.

After the step S22 It will be in one step S26 determines if the collision risk is above a predetermined threshold.

In the positive case (step S26 : Yes) the process goes to a step S28 in which it is determined whether the vehicle speed of the ego vehicle H Below a predetermined threshold, for example, 35 km / h, to prevent false-positive reactions.

In the positive case (step S28 : Yes) the process goes to a step S30 in which it is determined whether the steering wheel angle of the ego vehicle H above a predetermined threshold, for example above 25 °.

In the positive case (step S30 : Yes) the process goes to a step S32 in which an autonomous emergency braking (AEB) is performed. The process then returns to step S12 back and the subsequent steps are again performed as described above.

If the answer to one of the steps S26 . S28 . S30 is negative (step S26 : No or step S28 : No or step S30 : No) the process goes to a step S34 in which it is checked whether AEB is currently being applied.

In the positive case (step S34 : Yes) will AEB in one step S36 stopped and the process returns to step S12 back. In the negative case (step S34 : No) the process returns immediately S12 back.

With reference now to 4 the route prediction subroutine will be described in more detail.

After entering the route prediction subroutine in step S14 has occurred, it is in one step S40 determines if the ego vehicle H is close to a crossroads.

If the ego vehicle H near to a crossroads (step S40 : Yes) the process goes to a step S42 in which it is determined whether the steering wheel angle is below a threshold for the first time.

In the positive case (step S42 : Yes) the process goes to a step S44 in which the vehicle-fixed coordinate system of the ego vehicle is frozen in an intersection-fixed coordinate system.

The step S44 Following, the process goes to step S46 in which all relevant oncoming vehicles are detected and optionally, each oncoming vehicle is given an individual ID.

The same step S46 will also be executed when the answer to step S42 is negative (step S42 : No), since it is assumed that the vehicle-fixed coordinate system of the ego vehicle is already in an intersection-fixed coordinate system in an earlier call of the path prediction subroutine S14 has been frozen.

Then, in one step S48 , map data and relevant oncoming vehicles, which are in step S46 entered into the frozen coordinate system.

Next is at least one possible path of the ego vehicle H to perform the lane-crossing turn maneuver in step S50 estimated and at least one possible way is for each oncoming vehicle O to drive through the intersection in one step S52 valued. A detailed description of how this way of estimating the steps S50 and S52 will be explained below with reference to 7 are given.

Regarding 7 For example, the possible path of the ego vehicle H may be a pre-turn section S1 , a curved turn section S2 and a post-turn section S3 include. In a basic model, the pre-bending section S1 and the post-turn section S3 be approximated by straight lines while the turn section S2 can be approximated by a circular arc, the arc being the angle X between the track L1 on which the ego vehicle H the intersection approaches, and the track L2 covering on which the ego vehicle H the intersection leaves.

The way of the oncoming vehicle O can be approximated by a straight line.

In a subsequent step S54 become possible trajectories of the ego vehicle H calculated to perform the lane-crossing turn maneuver. It is clear that a trajectory is the sequence of positions of the ego vehicle H describes for a plurality of times. In other words, even if the ego vehicle H Moving along one and the same path, trajectories of the ego vehicle may be different when the ego vehicle H moved according to different speed profiles. As an example shows 6 a plurality of typical no-stop velocity profiles, that is, velocity profiles for performing a lane-crossing turn maneuver without the vehicle stopping at the intersection, in a graph showing vehicle speed (in km / h) of the ego vehicle H in the Y Axis relative to the distance (in m) of the ego vehicle H to the critical zone C (please refer 7 ) of the crossing in the X Axis shows. These speed profiles can be determined during test drives, for example.

Then the process goes to a step P.56 in which the collision rate D16 is determined. In particular, it checks which of the trajectories of the ego vehicle H in a collision with an oncoming vehicle O and the collision rate is determined as the number of trajectories that result in a collision divided by the total number of trajectories considered. $$KOllisiOnstrate=\frac{\text{Number of ego}-\text{vehicle}-\text{trajectories}\text{which result in a collision}}{\text{Number of all considered ego}-\text{vehicle}-\text{trajectories}}$$

A more refined approach to determining the collision rate will be discussed below with reference to FIG 8th to be discribed.

After the step P.56 or if the answer to the step S40 Is "No", the process returns to the main routine of 3 in one step S58 back.

• - eine Geschwindigkeit des Ego-Fahrzeugs H;
• - eine Geschwindigkeitsänderung des Ego-Fahrzeugs H;
• - eine Beschleunigung des Ego-Fahrzeugs H;
• - eine Beschleunigungsänderung des Ego-Fahrzeugs H;
• - eine Bremsdruckänderung des Ego-Fahrzeugs H;
• - eine Zeit des Ego-Fahrzeugs H, um mit der erfassten Verzögerung anzuhalten;
• - eine Distanz zu einer Konfliktzone;
• - eine Distanz zu einer Konfliktzone nach einem Anhalten mit einer erfassten Verzögerung;
• - eine Distanz zu einer Konfliktzone nach einem Anhalten mit einem komfortablen Bremsen;
• - einen Lenkwinkel;
• - eine Lenkwinkeländerung.

With reference now to 5 begins the maneuver prediction subroutine S18 with step S60 in which features are calculated based on the input data. These features may include, for example:

• - a speed of the ego vehicle H ;
• a speed change of the ego vehicle H ;
• - an acceleration of the ego vehicle H ;
• an acceleration change of the ego vehicle H ;
• - A brake pressure change of the ego vehicle H ;
• - a time of the ego vehicle H to stop with the detected delay;
• - a distance to a conflict zone;
• a distance to a conflict zone after stopping with a detected delay;
• a distance to a conflict zone after stopping with a comfortable braking;
• a steering angle;
• a steering angle change.

The calculated features then become a plurality of trained decision trees in one step S62 in which a series of yes / no questions are asked and answered, leading to a prediction of the intention of the driver of the ego vehicle H Perform the turn-by-turn turn maneuver without stopping at the intersection. Each of the trained decision trees outputs a probability value. A dialing unit combines all such probabilities of all decision trees into a so-called "no-stop probability" D20 ,

Then, the subroutine returns to the main routine of 3 in one step S64 back.

With reference now to 8th becomes an even more refined approach to determining the collision rate D16 are given:

Due to the fact that vehicles are not points, the size of the vehicles, at least in their width and length, must be taken into account when determining the risk of collision. According to the present invention, this is done by a three-circle representation of the vehicles which are in 8th is shown performed to reduce the required computation time.

According to 8th become the ego vehicle H or the oncoming vehicle O represented by first and second sets of three circles, respectively.

The first set of three smaller circles CS , which are shown as dashed lines, is adapted to represent the actual size of the respective vehicle. The position of the center of each circle may be at a longitudinal center plane ML be positioned of the respective vehicle, that is, a plane which divides the width of the vehicle into two equal left and right parts. The position of the center of the first and / or third circle may be such that an area extending beyond the front of the vehicle is substantially equal to an area which is not covered by the circle at the corners of the vehicle is covered. The position of the center of the second circle may be such that it is at a transverse center plane MT the corresponding vehicle is positioned, that is, a plane which divides the length of the vehicle into two equal front and rear parts.

The second set of three larger circles CL , which are shown as dotted lines, is arranged such that the total area of the three circles covers the entire vehicle. The position of the center of each circle of the second set of larger circles CL can be equal to the centers of the circles of the first set of smaller circles CS be positioned.

When the ego vehicle performs a turn-off turn maneuver, as in 8th shown, it checks if the sentences of smaller circles (first sentences CS ) of both vehicles overlap. If so, it is concluded that the vehicles will definitely collide, that is with a probability of 100%. Therefore, such collisions with a weighting factor of 1.0 are considered.

In order to be able to take data inaccuracies into account, it is further checked whether the sets of larger circles (second sentences CL ) of both vehicles overlap, and whether a set of small circles CS a vehicle with a set of big circles CL of the other vehicle overlaps. An overlap is considered to indicate a certain collision probability. This is considered by assigning weighting factors to the different types of overlaps. For example, a collision of two sets of larger circles CL with a factor of 0.6, which represents a small probability of collision, whereas a collision of a set of smaller circles CS with a set of larger circles CL can be weighted by a factor of 0.8, which represents a high probability of collision.

wobei gemäß dem oben erwähnten Beispiel Σc_small die Anzahl von Kollisionen/Überlappungen zwischen kleinen Kreisen ist, Σc_{small_large} die Anzahl von Kollisionen/Überlappungen zwischen einem kleinen Kreis und einem großen Kreis ist, Σc_large die Anzahl von Kollisionen/Überlappungen zwischen großen Kreisen ist, K_small der Gewichtungsfaktor für Kollisionen/Überlappungen zwischen kleinen Kreisen ist, zum Beispiel 1,0, K_{small_large} der Gewichtungsfaktor für Kollisionen/Überlappungen zwischen kleinen Kreisen und großen Kreisen ist, zum Beispiel 0,8, und K_large der Gewichtungsfaktor für Kollisionen/Überlappungen zwischen großen Kreisen ist, zum Beispiel 0,6.The weighted collision rate can then be calculated by the following formula: $$KOllisiOnstrate=\frac{K_{small}*{\displaystyle \Sigma c_{small}+K_{small\_larGe}*{\displaystyle \Sigma c_{small\_larGe}+K_{larGe}*{\displaystyle \Sigma c_{larGe}}}}}{\text{Number of all variations of the ego}-\text{vehicle}-\text{trajectories}}$$

where, according to the above-mentioned example, Σc _{small is} the number of collisions / overlaps between small circles, Σc _{small_large is} the number of collisions / overlaps between a small circle and a large circle, Σc _{large is} the number of collisions / overlaps between large circles, K _{small is} the weighting factor for collisions / overlaps between small circles, for example 1.0, K _{small_large is} the weighting factor for collisions / overlaps between small circles and large circles, for example 0.8, and K _{large is} the weighting factor for collisions / overlaps between large circles is, for example, 0.6.

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 patent literature

• WO 2004/068165 A [0004, 0005]

Claims (10)

A collision risk predicting unit (112) as a part of a vehicle driving support apparatus (100), wherein the vehicle driving support apparatus (100) is adapted to be mounted in an ego vehicle (H) and a vehicle Assisting the ego vehicle driver in avoiding a collision with another vehicle (O) when performing a lane-crossing turn maneuver, the collision risk prediction unit (112) having a collision rate determining subunit (124) and a probability determining subunit; that a vehicle does not stop, (126), wherein the collision rate determination subunit (124) is adapted to determine a collision rate (D16) as the percentage of those vehicle trajectories of the ego vehicle (H) from a predetermined plurality of vehicle trajectories in a collision with the other vehicle (O), the determination subunit for a probability (e) that a vehicle does not stop, (126) is arranged to determine a probability value that a vehicle does not stop (D20) as the probability that the ego vehicle (H) will continue the lane-crossing turn maneuver without stopping wherein the collision risk prediction unit (124) is arranged to determine a collision risk prediction value (D24) by multiplying the collision rate (D16) by the probability that a vehicle is not stopping (D20). Collision risk prediction unit after Claim 1 characterized in that the collision rate determining subunit (124) is further configured to determine the collision rate (D16) based on a vehicle trajectory, the vehicle trajectory being a predetermined lane crossing lane traversing a predetermined speed profile. Collision risk prediction unit after Claim 1 or 2 characterized in that the ego vehicle (H) and the oncoming vehicle (O) are represented by a group of three circles (CS), a first circle representing the front of the respective vehicle, a second circle which represents the center of the respective vehicle and a third circle, which represents the rear of the respective vehicle. Collision Risk Prediction Unit (112) after Claim 3 characterized in that the ego vehicle (H) and the oncoming vehicle (O) are additionally represented by a further group of three circles (CL), namely a first circle representing the front of the respective vehicle, a second circle , which represents the center of the respective vehicle, and a third circle, which represents the rear of the respective vehicle. Collision risk prediction unit after Claim 4 characterized in that the collision rate determining subunit (124) is further configured to collision between the ego vehicle (H) and the oncoming vehicle (O) based on an overlap of the smaller circle representations (CS) for both vehicles (H, O) based on an overlap of the larger circle representations (CL) for both vehicles (H, O) and on the overlap of the smaller circle representation (CS) for one of the vehicles (H or O) with the larger circle representation (CL) for the other one of the vehicles (O or H). Collision risk prediction unit after Claim 5 characterized in that the collision rate determining subunit (124) is adapted to collisions determined based on an overlap of the larger circle representations (CL) for both vehicles (H, O) with a lower weighting factor (K _large ) as collisions, which has been determined on the basis of an overlap of the smaller circle representation (CS) for one of the vehicles (H, O) with the larger circle representation (CL) for the respective other of the vehicles (O, H), and on the basis an overlap of the smaller circle representation (CS) for one of the vehicles (H, O) with the larger circle representation (CL) for the respective other of the vehicles (O, H) with a lower weighting factor (K _{small_large} ) than collisions based on an overlap of the smaller circle representations (CS) for both vehicles (H, O) have been determined. Collision risk prediction unit according to one of Claims 1 to 6 characterized in that the likelihood that a vehicle is not stopping determines (126) a plurality of trained decision tree sub-units, each configured to output an individual likelihood value that a vehicle is not stopping, and a dial sub-unit, which is adapted to the probability value that a vehicle does not stop (D20) on Basis of the plurality of individual probability values that a vehicle does not stop to determine. Collision risk prediction unit after Claim 7 characterized in that each of the trained decision tree sub-units takes into account at least one of the following parameters: a speed of the ego vehicle (H); a speed change of the ego vehicle (H); an acceleration of the ego vehicle (H); an acceleration change of the ego vehicle (H); a brake pressure change of the ego vehicle (H); a time of the ego vehicle (H) to stop with the detected delay; - a distance to a conflict zone; a distance to a conflict zone after stopping with a detected delay; a distance to a conflict zone after stopping with a comfortable braking; a steering angle; and a steering angle change. Collision risk prediction unit according to one of Claims 1 to 8th characterized in that the collision risk prediction unit (124) is adapted to consider environmental structure information. Collision risk prediction unit after Claim 9 characterized in that the environmental structure information is provided by a navigation system and / or by an e-horizon system and / or by at least one sensor (104).

Images