DE102012009297A1 - Method for assisting rider when feeding e.g. vehicle, involves proving information, warning and automatic engagement, which results during risk of collision and/or secondary collision with highest priority in priority list - Google Patents

Abstract

The method involves informing a priority list and driver preferences to a driver of a vehicle in accordance with a time of a desired system interference and/or a determined driver condition of the driver in a number of escalation steps by driver instructions, warning and/or supported by an automatic intervention in a longitudinal and/or lateral control of the vehicle. An information, warning and an automatic engagement, which results during a risk of collision and/or secondary collision is provided with a highest priority in the priority list.

Description

The invention relates to a method for assisting a driver when driving a vehicle, driver information being output as a function of a predicted future potential risk of collision and / or subsequent collision risk between the vehicle and other road users within the vehicle.

In the not yet published DE 10 2011 120 117.7 A method for optically signaling a potential danger of a collision between a vehicle and another road user for a driver of a vehicle is described, wherein a degree of distraction of the driver is determined based on results of driver observation using an in-vehicle camera and operator actions. Using sensor data, a state of the vehicle as well as a location and movement of other road users are determined, wherein, based on the state of the vehicle, the location, and the movement of the other road users, a prognosis of a future potential danger of collision between the vehicle and another Road users is created. If a detected degree of distraction of the driver and a predicted degree of potential danger of a collision between the vehicle and another road user exceed a predefinable threshold, a perceptible warning to the driver is issued. The warning is configured to indicate to the driver a direction in which there is another road user relative to the vehicle ahead, for the future potential danger of a collision between the vehicle and the road user. A type and intensity of the warning depends on the degree of potential danger of the collision and a degree of distraction of the driver.

The invention has for its object to provide a comparison with the prior art improved method for assisting a driver when driving a vehicle.

The object is achieved by a method having the features specified in claim 1.

Advantageous embodiments of the invention are the subject of the dependent claims.

In a method for assisting a driver in driving a vehicle, driver hints are output in dependence on a predicted future potential collision risk and / or subsequent collision risk between the vehicle and other road users within the vehicle.

The vehicle is one of the road users.

A probabilistic risk assessment in space-time is carried out cognitively using competing hypotheses between a vehicle and other road users. A potential risk of collision is evaluated on the basis of probabilistically recognized maneuver intentions of the vehicle and the other road users. At the same time, the respective real risk of collision and / or consequential risk of collision is determined in corresponding conflict areas, with a risk assessment being carried out for each situation hypothesis on the basis of logical context rules, including priority rules. Parallel to the risk assessment, a probabilistic procedure for determining the real and acute danger of collision is carried out with a determination of the range of motion. In this case, a large number of mobile trajectories of the vehicle and of the other road users (V1 to Vn) is predicted on the basis of their statistical control profiles over a time horizon of 2 seconds.

Based on the trajectories, several competing hypotheses of the situation between the vehicle and all other relevant road users are determined, and the available range of motion is determined with the appropriate probability. The probabilistic events thus obtained from the two methods, risk assessment and range of motion, are hierarchically arranged in a priority list as a function of the respective risk of collision and / or consequential risk of collision. The driver of the vehicle is informed, warned and / or assisted by means of an automatic intervention in a longitudinal and / or lateral control of the vehicle in a plurality of escalation levels depending on the priority list and an ascertained driver state of the driver, the information, the warning and the automatic intervention takes place for the situation which has a risk of collision and / or secondary collision risk with the highest priority in the priority list and a consequent highest criticality.

The method allows in a particularly advantageous manner a parallel evaluation of competing situation hypotheses between all relevant objects of a situation. In spite of different dynamic differences, the method is suitable for use in transverse, counter and longitudinal traffic, since the risk assessment of a particular situation takes place under logical context rules, which in particular are priority and traffic rules, speed limits, display states of traffic sign systems and variable traffic signs as well as a topography and topology of a corresponding digital road map. The context rules are integrated in so-called object-oriented Bayesian networks, whereby the hypotheses of the situation are determined and evaluated in particular on the basis of these object-oriented Bayesian networks. In addition, due to the probabilistically determined driver intentions for the risk assessment and based on the risk of collision with movement latitude and the generation of the probabilistic and hierarchically ordered priority list of collision hazards and / or consequential collision hazards and the combination of this priority list with the driver state can be avoided that unnecessary and the driver irritating information and warnings are issued.

Embodiments of the invention are explained in more detail below with reference to drawings.

Showing:

1 schematically a first traffic situation between several road users with their maneuver options in a road intersection area,

2A schematically a second traffic situation between two oncoming vehicles as well as sets of movement hypothesis trajectories of the vehicles,

2 B schematically a hazard probability distribution as a function of a longitudinal acceleration and a steering rate of a first vehicle according to 2A .

2C schematically a hazard probability distribution as a function of a longitudinal acceleration and a steering rate of a second vehicle according to 2A .

3A schematically a third traffic situation between two oncoming vehicles as well as sets of movement hypothesis trajectories of the vehicles,

3B schematically a hazard probability distribution as a function of a longitudinal acceleration and a steering rate of a first vehicle according to 3A .

3C schematically a hazard probability distribution as a function of a longitudinal acceleration and a steering rate of a second vehicle according to 3A .

4A schematically a fourth traffic situation between two vehicles traveling in the same direction next to each other as well as sets of movement hypothesis trajectories of the vehicles,

4B schematically a hazard probability distribution as a function of a longitudinal acceleration and a steering rate of a first vehicle according to 4A .

4C schematically a hazard probability distribution as a function of a longitudinal acceleration and a steering rate of a second vehicle according to 4A

5 schematically a determination of driver intentions from different data,

6 schematically a fifth traffic situation with several vehicles,

7 1 schematically shows a risk assessment procedure for determining a collision probability between two vehicles and an output of information in a first escalation stage and a first warning in a second escalation stage;

8th schematically a process flow of a method for assisting a driver in driving a vehicle and

9 schematically a procedure for determining a collision probability between two vehicles and information, two preparatory warnings and a final warning of an intervention in a longitudinal or lateral control of a vehicle as a function of the probability of collision in several escalation stages.

Corresponding parts are provided in all figures with the same reference numerals.

In 1 is a first traffic situation between several road users V1 to V12 shown with their maneuver options in a road intersection area. For all situations described in the following figures, one of the road users represented represents the own vehicle in the sense of the method according to the invention.

The road users V1 to V12 include vehicles (road users V1 to V4), endangered or endangered pedestrians (road users V5 to V8) and cyclists (road users V9 to V12). The vehicles (road users V1 to V4) include motorcycles, passenger cars, trucks, buses and vans. The Cyclists in particular move on cycle paths next to the road.

The intersection area represents a simple "4-way" intersection topology and has one access lane in each direction. The method described below is generic and extendable to intersections of arbitrary topology and geometry, as appropriate contextual information is extracted from a digital roadmap of the intersection and integrated into a structure of object oriented Bayesian networks to determine appropriate maneuvering options.

When determining a size of dangers of the first traffic situation, a cognitive evaluation of potential and real dangers of the current traffic situation is carried out by means of a probability-based method, wherein in the present exemplary embodiment an object-oriented Bayes network is used in a probability-based method.

The determination of one in the 7 and 9 collision danger DA between the traffic participants V1 to V12 shown in more detail by means of the not yet published DE 10 2011 113 019 described method, here with the term "risk assessment in space-time" referenced.

The driving options of the vehicles and the movement options of the unprotected road users are generally identified as maneuver options or possible maneuver tracks. The maneuver options are cognitively identified by contextual information from the digital map, motion patterns, and motion characteristics. A trajectory bundle, synonymous with the set of motion hypothesis trajectories, is predicted using physical models. For example, for this prognosis, the "one-track model" applied to the set of statistically determined control variables, such as longitudinal acceleration and steering rate, is used. The maneuver options, on the other hand, define all possible movements on a road type. For example, the possible maneuvering options for the vulnerable road users are before an intersection: waiting at a traffic light, crossing at a pedestrian crossing or lane crossing from X to Y roadside. The possible maneuver options for the vehicles before an intersection are: waiting at the traffic light, at the stop line or at the sight line, or driving in the direction left / right / straight as defined by the intersection topology. On a multi-lane road, the possible maneuver options for the own or object vehicle are: track following, luffing from the currently occupied lane to left / right, or lapping into the adjacent lane to left / right, following object, threading into the faster traffic flow of a feeder track or unthreading from the faster traffic flow on a departure lane.

The determination of the maneuvering options takes place on the basis of intentions of the road users V1 to V12, which in the illustrated road intersection area in 1 possible are. In this case, possible intended maneuver tracks (for the road intersection area) MT1 to MT12 of the road users V1 to V12 are taken into account, the maneuver tracks MT1 to MT12 being generated from past, current and predicted trajectories of the respective road users V1 to V12. The predicted trajectories are presented in a similar way as the conflict representation trajectories of a crossing, ie in accordance with the applicable rules for the traffic engineering design of phase interim times of traffic lights. The representative maneuver tracks are used as representative lanes when turning or crossing an intersection. In addition, the representative maneuver tracks are used as a reference for tracking the vehicles while tracking through the intersection area. Furthermore, position and movement relationships of the road users V1 to V12 are taken into account. These relationships are also included in the digital road map.

The determination of the maneuvering options on other road types with longitudinal and oncoming traffic also takes place on the basis of possible intentions of the road users V1 to V12 on the corresponding roadway types. The driver's intentions are based on different data, as in 5 represented, cognitively recognized. The cognitive recognition of the driver's intentions FA of the drivers of the vehicles takes place on the basis of position data of the road users V1, V2, movement data of the road users V1, V2 and environmental information of the road users V1, V2. The driver intent recognition is performed by means of a probability-based method, wherein in the present exemplary embodiment an object-oriented Bayes network is used.

The thus-recognized driving intentions are combined in pairs to determine all possible intersections of the current maneuver intentions per road user.

In particular, as in 7 In a first step of the cognitive evaluation, a probabilistic interpretation of the maneuver intentions of all participants in the traffic situation is presented, in a second step a probabilistic hypothesis estimation of basic hypotheses for the estimation of mutual intersections of the intended Maneuvers of road users V1 to V12 and of the occupancy of the same conflict areas by road users V1 to V12. The second step calculates the potential risk of collision in the room. In a third step, the simultaneous occupancy of the conflict areas is additionally taken into account in order to derive a probabilistic hypothesis estimation of the risk of collision DA and / or consequential risk of collision for the road users V1 to V12 in relative movement to their own vehicle.

The probabilistic hypothesis estimation described above becomes particularly dependent on traffic rules, display conditions of traffic lights and variable traffic signs, absolute positions of the traffic users V1 to V12, relative positions of the traffic participants V1 to V12 to each other and of intended maneuver tracks MT1 to MT12 of the road users V1 to V12 and as a result delivers the context-related risk assessment in space-time.

The determination of the real risk of collision with freedom of movement or the acute risk of collision without freedom of movement, ie with a need for extreme maneuvers, is carried out by means of the method of not yet published DE 10 2012 005 272 , which is referenced here by the term "Feel-Safe-Zone". This will track the collision avoidance and / or sharpening.

2A shows a second situation between two oncoming traffic participants V1, V2, which are designed as in mutually relative movement vehicles. Furthermore, the traffic participants V1, V2 associated amounts M1, M2 of movement hypothesis trajectories T1 ₁ to Tm ₁ , T1 ₂ to Tn ₂ are shown. The road users V1, V2 move, for example, within a closed town at speeds between 30 km / h and 50 km / h in a crossroads area.

To determine one in the 2 B and 2C Danger probability P (C) of the situation between the road users V1, V2, which are shown in greater detail, are predicted for future movement trajectories of the road users V1, V2 by generating the movement hypothesis trajectories T1 ₁ to Tm ₁ , T1 ₂ to Tn _{2 of} the road users V1, V2. The danger probability P (C) indicates a probability with which a collision K between the road users V1, V2 takes place and corresponds to the determined in the second process stage "feel-safe zone" with the available control variables, acceleration and steering wheel rate, for the currently analyzed sitivation.

The determination of the position data for both procedures ("risk assessment in space-time" DE 10 2011 113 019 and "Feel Safe Zone" DE 10 2012 005 272 ) is carried out by means of localization methods, wherein in the present embodiment, a global positioning system, for example dGPS (differential GPS), or a digital map with landmarks is used. These localization methods provide a true-to-track localization of the vehicles (in the centimeter range with dGPS up to approx. 30 cm with the digital map). The accuracy for locating vulnerable road users is on the order of 30 cm, while the accuracy of their intentional recognition is in the centimeter range to detect movement patterns. This accuracy is important for the detection performance of a reliable V2X (vehicle-to-vehicle and vehicle-to-infrastructure) communication solution for situational interpretation and risk assessment for increasing traffic safety as well as for driver acceptance of the driver assistance system. Other alternative technologies for on-track localization, in addition to the use of digital maps and landmarks, are tightly-coupled GNSS / INS, Global Navigation Satellite System (GNSS) cooperative triangulation with two reference transmitters. These ensure an accuracy of at least 0.3 m, which laterally is sufficient for the track-accurate localization.

The determination of the movement data also takes place on the basis of the global positioning system and / or by means of the vehicle's own sensors, for example by means of at least one speed sensor, steering rate sensor, yaw rate sensor, acceleration sensor, an environment detection unit and / or a compass.

The environment information of the road users V1, V2 are detected by means of the at least one environment detection unit. This environment detection unit is an image acquisition unit, for example an in-vehicle and / or one or more infrastructure-bound camera / s, which has a detection range of, in particular, 360 ° after the data fusion. Alternatively or additionally, in particular for realizing missing data for generating the coverage of 360 °, the environmental information from map data of a digital road map, from communicated data from an ITS-standardized Cooperative Awarness Message, a vehicle-to-infrastructure communication and / or a Road user-to-vehicle communication or vehicle-to-vehicle communication. From the environment information is in turn derived, whether in the environment of road users V1, V2 are danger spots and / or areas of conflict.

To determine the driver's intentions, the road users V1, V2 are located on their lane. In this case, a relative speed of a preceding road user V1, V2 on the same lane together with a relative speed of a following road user V1, V2 on the same lane and a relative speed of an oncoming, parallel driving or overtaking road user V1, V2 on an adjacent lane or the same lane. Also, the driver intentions are derived from motion patterns and lights, such as from activation of a direction indicator.

To identify the intentions of unprotected road users, the vulnerable or endangered pedestrians (road users V5 to V8) and cyclists (road users V9 to V12) are located on the appropriate segment of the road or intersection. Classification algorithms determine their movement characteristics, such as direction of movement, orientation and relative speed, derived from the locally fused data and / or through V2X-communicated data. These features are used in combination with the traffic rules (traffic signs and traffic lights) as input data for the intentional identification of unprotected road users. From this, their possible movement options, generally called maneuver options, are derived.

The possible mutual points of intersection of the maneuver options represent positions of potential collisions K between the traffic participants V1 ... Vn expected in the conflict areas and considered in pairs, so that the potential collision risk K and / or consequential risk of collision between the road users V1 ... Vn be determined in pairs.

The risk assessment is derived from the cognitive combination of events:
Potential collision risk, simultaneous occupation of the conflict areas and compliance with priority rules. It is taken into account as an indication of the real risk of collision and / or consequential risk of collision in space-time.

In a second stage "Feel-Safe-Zone" following the first stage of the process "Risk Assessment in Space-Time", potential collision-free movement hypothesis-trajectory pairs of road users V1, V2 are determined and evaluated on the basis of the determined real risk of collision K. The possible collision-free movement hypothesis trajectory pairs of road users V1, V12 represent maneuvering options of road users V1, V12 for the respective situation, which can be carried out without collision K. These are used as input variables for collision-free roadway planning.

When assessing the collision-free movement hypothesis trajectory pairs, the respective ranges of motion between the road users V1, V12 are determined, and the danger probability P (C) is determined as a function of a size of the respective range of motion. When evaluating the range of motion is an in 9 shown driver state FZ and / or a driver presetting taken into account, wherein for determining the driver state FZ an attention level, a degree of distraction, a level of fatigue, stress and / or vital parameters of the driver are detected. The driver presetting includes the general request of the driver independently or with the assistance of the driver assistance system of a currently detected situation in which a range of motion is present to meet. The system distinguishes whether a range of motion, without or with extreme maneuvers, exists. If the calculated range of motion allows another road user to pass (in a comfortable and secure manner) and the driver preset includes the desire for autonomous control, the driver is given the opportunity to perform a correction maneuver himself after the second system warning has been issued. If the driver acts correctly, the situation is defused, with no need for system intervention. If the driver does not act correctly, ie the available range of motion is not used, or if the default setting is "system support in the event of a collision hazard" or if there is no room for maneuver, a third warning is issued followed by a system intervention that prevents or mitigates a collision.

The determination of the probability of danger P (C) and the range of motion, here referred to as "Feel-Safe-Zone", is determined on the basis of a stochastic model according to the DE 10 2011 106 176 A1 described method and / or according to the not yet published DE 10 2012 005 272 carried out. Thus, the hazard assessment is based on a prognosis of all possible drivable and controllable trajectories for road users V1 to V12. In particular, the forecast has a lead time of 2 s. In other words, the predicted trajectories are known 2 s before the actual event occurs.

This means - the combination of the two methods - "risk assessment in space-time" and the calculation of the probability of danger P (C) and available range of motion (ie "feel-safe zone") - provides the overall context-dependent derivation of the real risk of collision with the available one Movement scope for collision avoidance. Both methods - risk assessment of the potential and real risk of collision and "feel-safe zone" to determine the real and acute risk of collision with the available range of motion - run parallel to the situation analysis in the transverse, longitudinal and oncoming traffic. Both methods provide the triggers for the four escalation levels of the system at various times: information, warnings (W1 to W3) and / or active intervention / intervention in the longitudinal or lateral dynamics.

When determining the danger probability P (C), an action of both road users V1, V2 in oncoming traffic, in particular within closed localities in arbitrary speed ranges, is taken into account in the illustrated situation. Furthermore, overtaking maneuvers with and without oncoming traffic as well as shearing and shearing, threading / unthreading, or a lane change are taken into account, both within built-up areas and outside built-up areas on motorways, motor roads and highways. Curved road profiles are also taken into account, the curvature of the road profiles being derived, in particular, from the yaw rate of the respective road user V1, V2 depicted as vehicle.

2 B shows a hazard probability distribution as a function of a longitudinal acceleration a and a steering rate l of the first road user V1. The hazard probability distribution is subdivided into four danger areas B1 to B4, the danger probability P (C) increasing from the first danger area B1 to the fourth danger area B4. The risk probability P (C) in the first danger zone B1 is in particular up to 25%, in the second danger zone B2 25% to 50%, in the third danger zone B3 50% to 75% and in the fourth danger zone B4 75% to 100%.

It can be seen from the hazard probability distribution that with increasing acceleration a and an increasing steering rate l in a left direction the danger probability P (C) of a collision K with the second road user V2 increases.

In 2C the probability distribution of the same situation for the second road user V2 is shown. Again, the hazard probability distribution is divided into four hazard areas B1 to B4. It can be seen that with increasing acceleration a and an increasing steering rate l in a left direction, the danger probability P (C) of a collision K with the first road user V1 increases.

In 3A is a third traffic situation between two oncoming traffic participants V1, V2 shown. Unlike the in 2A shown second traffic situation, the road users move V1, V2 on a two-lane road towards each other.

The 3B and 3C show the probability distribution of the road users V1, V2 as a function of the longitudinal acceleration a and the steering rate I, with increasing acceleration a and an increasing steering rate l of the first road user V1 in a left direction, the danger probability P (C) of a collision K with the second road user V2 increases and increases with increasing acceleration a and an increasing steering rate l of the second road user V2 in a left direction, the danger probability P (C) of a collision K with the first road user V1.

4A shows a fourth situation between two in the same direction side by side moving road users V1, V2, ie, in relative motion to each other vehicles, and the road users V1, V2 associated amounts M1, M2 of movement hypothesis trajectories T1 ₁ to Tm ₁ , T1 ₂ to Tn ₂ . The road users V1, V2 move for example on a highway, motor road or country road. Such a situation shown results, for example, in Ausscher- and Einschermanövern, Ausfädel- and Einfädelmanövern at exits and driveways and lane change maneuvers on the mentioned types of road, such lane change maneuvers in 6 are shown in more detail.

The determination of in the 4B and 4C Danger probability P (C) of this situation between the road users V1, V2 shown in more detail analogous to that in 2A illustrated situation.

In 4B a hazard probability distribution as a function of the longitudinal acceleration a and the steering rate l of the first traffic participant V1 is shown. The hazard probability distribution is subdivided into the four hazard areas B1 to B4, the danger probability P (C) increasing from the first danger area B1 to the fourth danger area B4. In this case, the danger probability P (C) in the first danger zone B1 is in particular up to 25%, second danger zone B2 25% to 50%, in the third danger zone B3 50% to 75% and in the fourth danger zone B4 75% to 100%.

It can be seen from the hazard probability distribution that with decreasing acceleration a, in particular with a negative acceleration a, and depending on an increasing steering rate l in a left direction, the danger probability P (C) of a collision K with the second road user V2 increases.

4C shows the hazard probability distribution of the same situation for the second road user V2. Again, the hazard probability distribution is divided into four hazard areas B1 to B4. It can be seen from the representation that the danger probability P (C) of a collision K with the first road user V1, V2 also increases with decreasing, in particular negative acceleration a, but with an increasing steering rate l in a right direction.

In 5 schematically a determination of driver intent FA is shown from different data. The driver's intentions FA include an intention of a respective driver to perform luffing and / or Ausschorgorgänge EA, threading and / or Ausfädelvorgängen EA, lane change operations S and overtaking U.

These driver intentions FA are in the inventive method for situation analysis of Einscher- and / or Ausschervorgänge EA of in 6 used in detail and each trained as a vehicle road users V1 to Vn used.

In this case, a lateral evidence, the trajectories of road users V1 to Vn and object-oriented dynamic grids dG probabilistically combined by means of the object-oriented Bayes network to determine the maneuver intentions of the driver. For this purpose, the lateral evidence is determined by the following features: distance to the lane marking and lateral velocity. The trajectory is calculated on the basis of the characteristics: time for crossing the lane marking, maximum utilized acceleration and toe angle error. The assignment of the dynamic grid dG is determined by the occupancy time of a cell and by the distance from the current position of the road user to an entry into the cell or until it leaves the cell. An object-oriented dynamic grid dG is also in 6 shown in more detail.

Furthermore, as context information from a digital road map of a road segment SS1 to SSu on which the respective traffic participant V1 to Vn is located, signals from not shown vehicle-side sensors of the road users V1 to Vn and from data of a vehicle-to-vehicle communication and / or a vehicle-to-infrastructure communication between the road users V1 to Vn itself and / or between the road users V1 to Vn and an infrastructure in 7 shown movement state BZ1 to BZn the road users V1 to Vn, a control state SZ1 to SZn the road users V1 to Vn, a driver activity state FAZ1 to FAZn and event messages EN determined. The road segments SS1 to SSu are centered as dynamic allocation grids around a respective traffic participant V1 to Vn formed by the object-oriented dynamic grid dG and are also in 6 shown in more detail. The dynamic allocation grids move along with the road users V1 to Vn.

In the situation analysis, it is determined from the context information, the lateral evidence, the trajectories of the road users V1 to Vn, the object-oriented dynamic grids dG, the distances, the relative orientations and the relative positioning of the road users V1 to Vn whose maneuver intentions.

When determining the context information from the digital road map of the road segment SS1 to SSu, a topography and topology of the road segment SS1 to SSu, traffic signs, traffic rules and / or lane markers on an in 6 shown lane FS1 to FS4 used. The context information determined from the digital road map is shown as information 11 and used to determine the driver's intent FA. In particular, the information I1 also includes the context as to whether a respective adjacent traffic lane FS1 to FS4 is a luffing and / or discharge lane or threading and / or delisting lane or an oncoming traffic lane.

When determining the context information by means of the vehicle-mounted sensors, a respective environment of the road users V1 to Vn in a detection range of 360 °, i. H. in a panoramic view, captured. In this case, the further road users V1 to Vn present in the surroundings of the respective road user V1 to Vn are detected and their relative positions relative to one another are determined. From these context information positions POS1 to POSn of road users V1 to Vn are determined on their lane FS1 to FS4, which in turn are used to determine the driver's intent FA.

Furthermore, by means of on-board sensors and the vehicle-to-vehicle Communication and / or the vehicle-to-infrastructure communication between road users V1 to Vn and / or the infrastructure positions POS2 to POSn, determines a speed and orientation of the road users V1 to Vn with respect to their lane FS1 to FS4. The vehicle-to-vehicle communication and / or the vehicle-to-infrastructure communication thereby enable the detection of possibly missing information to realize the coverage of 360 ° and thus allow a robust fusion of the various data to determine the context information.

From these data, the movement states BZ1 to BZn of the individual road users V1 to Vn and relative variables which are used in the determination of the driver's intention FA are again determined.

The relative variables comprise, starting from a respective road user V1 to Vn, an associated first relative speed v1 _{rel of} a further road user V1 to Vn traveling ahead of the respective road user V1 to Vn on the same traffic lane FS1 to FS4 and a second relative speed v2 _rel behind the respective road user V1 to Vn on the same lane FS1 to FS4 driving another road user V1 to Vn or one of the respective road users V1 to Vn on a secondary lane oncoming, parallel driving or overtaking other road users V1 to Vn.

Furthermore, the relative quantities include a first relative orientation O1 _{rel of} the respective road user V1 to Vn to the on the same lane FS1 to FS4 ahead of the respective road users V1 to Vn driving other road users V1 to Vn and a second relative orientation O2 _rel of behind the respective Road users V1 to Vn on the same lane FS1 to FS4 driving other road users V1 to Vn or of the respective road users V1 to Vn on a secondary lane oncoming, parallel driving or overtaking other road users V1 to Vn.

The driver activity state FAZ1 to FAZn of the road users V1 to Vn is determined from an active route of a navigation system, activated direction indicators, an activated hazard warning light, a position of a brake pedal, a position of an accelerator pedal and / or a steering wheel rate and also taken into account in the determination of the driver's intent FA.

Furthermore, the control state SZ1 to SZn of the road users V1 to Vn is taken into account in the determination of the driver's intent FA, the respective control state SZ1 to SZn being determined from a steering angle, an acceleration and / or a deceleration of the road users V1 to Vn.

The event messages EN are likewise taken into account in the determination of the driver's intention FA and are determined from obstacles and unprotected road users V1 to Vn and their positions on the lanes FS1 to FS4, the obstacles and the unprotected road users V1 to Vn being connected to the vehicle's own sensors, the vehicle sensors. for vehicle communication and / or vehicle-to-infrastructure communication.

In 6 is a fifth traffic situation with several road users, V1 to Vn shown.

In the illustrated first traffic situation, all road users V1 to Vn move in the same direction to four lanes FS1 to FS4. Depending on the lane FS1 to FS4, different maneuver options are available for the road users V1 to Vn, the maneuvering options including the shearing and / or shunting operations EA, the lane changing operations S, overtaking operations U and / or straight-ahead driving operations G connected thereto. For clarity, not all maneuver options are provided with reference numerals.

In the straight-ahead driving G follows the respective road users V1 to Vn of the lane FS1 to FS4, in which it is located. Einschervorgänge are characterized by the fact that a road user V1 to Vn changes into an adjacent lane FS1 to FS4 and then follows the respective lane FS1 to FS4. Ausschervorgänge distinguished by the fact that a road user V1 to Vn after a straight-ahead driving G changes into an adjacent lane FS1 to FS4. An overtaking process U is characterized in that a road user V1 to Vn changes to a left adjacent traffic lane FS1 to FS4 after a lane change operation S, follows the respective traffic lane FS1 to FS4 and subsequently completes a lane change operation S in an adjacent right traffic lane FS1 to FS4. Overtaking from the right does not happen according to the traffic rules, but they are also taken into account because of the increased risk for a collision.

Threading operations are distinguished by the fact that a road user V1 to Vn after a straight-ahead travel process G onto a feeder road into an adjacent traffic lane FS1 to FS4, in which traffic travels faster, both within closed towns as well as outside built-up areas on highways, motor roads and country roads, changes.

Ausfädelvorgänge are characterized by the fact that a road user V1 to Vn after a straight-ahead driving G both within closed towns and outside built-up areas on highways, motor roads and highways to an adjacent departure lane FS1 to FS4 changes to reduce its speed, d. H. adapt to the streets with slower traffic.

Some of the above road types use the same lane for both threading and unthreading operations, which is associated with a significantly increased risk of collision.

In the situation analysis and related, in 7 shown risk assessment RB the illustrated fifth traffic situation are using the object-oriented Bayes network lateral evidence, trajectories of road users V1 to Vn, the object-oriented dynamic grid dG together with the in 5 illustrated and described contextual information, the driving intentions FA probabilistic determined.

Within the object-oriented dynamic grating dG, in particular vehicle data, such as a respective length, width and height of the road users V1 to Vn, their positions POS1 to POSn, an absolute distance of the road users V1 to Vn to each other, lateral accelerations of the road users V1 to Vn and traversing times until the crossing of lane markings of the lanes FS1 to FS4 determined. The lane markings may be virtual lane markers, which result, for example, from a traffic flow of different groups of road users V1 to Vn, for which purpose preferably the entire roadway is detected by means of a camera and the captured images are evaluated.

Furthermore, conflict areas K1 to Ku are determined in the road segments SS1 to SSu of the object-centered dynamic lattice dG, with time periods for driving on a new road segment 551 to SSu and thus a conflict area K1 to Ku and for leaving a road segment SS1 to SSu and thus a conflict area K1 to Ku are determined.

A cell size of the dynamic allocation grid, i. H. a size of the road segments SS1 to SSu, is selected as a function of the speed of the road users V1 to Vn and thus adapted to different traffic situations and possible hazards. Also, a trend of a partial occupancy of the adjacent road segments SS1 to SSu in the vicinity of a respective road user V1 to Vn is determined, the trend being used to check the plausibility of the respective driver intention FA. An increasing occupancy of the adjacent road segments SS1 to SSu affirms, for example, the driver's intention FA. That is, by means of the dynamic allocation grid, a probability for simultaneous occupancy of the road segments 551 to SSu of the dynamic grid dG is determined by a plurality of road users V1 to Vn.

The probability of simultaneous occupancy of the road segments 551 to SSu is determined on the basis of several features.

These features comprise a first relative time duration (dTTE) up to a simultaneous occupancy of the road segments 551 to SSu by a plurality of road users V1 to Vn, this first relative time duration (dTTE) being dependent on the respective speed of the road users V1 to Vn.

dTTE

= relative erste Zeitdauer,

TTE_V1

= Zeitdauer für Verkehrsteilnehmer V1 und

TTE_V2

= Zeitdauer für Verkehrsteilnehmer V2.

The relative first time duration (dTTE) results according to the following equation from the difference of the respective time periods (TTE = time to enter) to the occupancy of the road segments SS1 to SSu by way of example for road users V1 and V2: dTTE = (TTE _V1 - TTE _V2 ), (1) With:

DTTE

= relative first time duration,

TTE _V1

= Time duration for road users V1 and

TTE _V2

= Duration for road users V2.

The features further comprise a second relative time duration (dTTD) until the road users V1 to Vn leave the road segments SS1 to SSu, this second relative time duration (dTTD) being dependent on the respective speed of the road users V1 to Vn.

dTTD

= relative zweite Zeitdauer,

TTD_V1

= Zeitdauer für Verkehrsteilnehmer V1 und

TTD_V2

= Zeitdauer für Verkehrsteilnehmer V2.

The relative second time duration (dTTD) results according to the following equation from the difference of the respective time periods (TTD = time to disappear) up to leaving the road segments SS1 to SSu for the road users V1 and V2 by way of example: dTTD = (TTD _V1 - TTD _V2 ), (1) With:

dTTD

= relative second time duration,

TTD _V1

= Time duration for road users V1 and

TTD _V2

= Duration for road users V2.

In addition, the features include a relative distance between a reference vehicle, respectively, depending on the relative first time duration and the relative second time duration. H. own vehicle, and another road user V1 to Vn.

dS(TTE)

= relativer Abstand,

S(TTE_V1)

= Abstand bis zur Belegung für Verkehrsteilnehmer V1 und

S(TTE_V2)

= Abstand bis zur Belegung für Verkehrsteilnehmer V2.

The relative distance is given as a function of the first relative time duration (dTTE) as follows: dS (TTE) = (S (TTE _V1 ) -S (TTE _V2 )), (1) With:

dS (TTE)

= relative distance,

S (TTE _V1 )

= Distance to occupancy for road users V1 and

S (TTE _V2 )

= Distance to occupancy for road users V2.

dS(TTD)

= relativer Abstand,

S(TTD_V1)

= Abstand bis zum Verlassen für Verkehrsteilnehmer V1 und

S(TTD_V2)

= Abstand bis zum Verlassen für Verkehrsteilnehmer V2.

The relative distance is given as a function of the second relative time duration (dTTE) as follows: dS (TTD) = (S (TTD _V1 ) -S (TTD _V2 )), (1) With:

dS (TTD)

= relative distance,

S (TTD _V1 )

= Distance until leaving for road users V1 and

S (TTD _V2 )

= Distance until leaving for road users V2.

Furthermore, a degree of visibility to the other road users V1 to Vn is determined.

On the one hand, the movement states BZ1 to BZn of the road users V1 to Vn are determined from the above-mentioned data, the movement states BZ1 to BZn being from an occupancy of the conflict areas K1 to Ku, the time periods until leaving and driving on a road segment SS1 to SSu, the distance to the road segments SS1 to SSu, the dynamic size of the road segments SS1 to SSu, the relative distance between the road users V1 to Vn, the relative orientation O1 _rel , O2 _rel the road users V1 to Vn each other, a relative differential speed between the traffic flow on the lanes FS1 to FS4 and to the respective other road users V1 to Vn in the environment as well as from a time duration and a distance to a licker-in / lane-lane mark or to a threading / unthreading lane mark.

The control states SZ1 to SZn and driver activity state FAZ1 to FAZn of the road users V1 to Vn are determined from the steering angle, the acceleration and / or deceleration of the road users V1 to Vn, a blind spot, d. H. the visibility of the other road users V1 to Vn, from the active route of the navigation system, activated direction indicators, the activated hazard warning light, the position of the brake pedal, the position of the accelerator pedal and / or the steering wheel determined.

The determination of the characteristics of the direction indicator, the vision concealment, the steering, the pedals and the warning light is done in a situation-dependent trend analysis with time.

In addition, the event messages EN are determined.

The risk assessment RB is carried out by determining possible intersections of maneuver pairs of the road users V1 to Vn and of simultaneous assignments of the conflict areas K1 to Ku. From these intersections and the simultaneous assignments, an expected risk of collision DA is determined spatially and temporally.

The risk of collision DA and / or consequential risk of collision are determined in sections for all possible maneuver pairs of road users V1 to Vn with a common intersection point, the maneuver pairs being determined from the driver's intentions FA of the road users V1 to Vn.

In the risk assessment RB, all characteristics of the road users V1 to Vn, d. H. the vehicle data, the position data, the movement states BZ1 to BZn, the control states SZ1 to SZn, the driver activity states FAZ1 to FAZn, and the event messages EN are merged with each other.

The collision risk DA in the risk assessment RB is determined by associating danger points of the digital road map with the conflict areas K1 to Ku. The association takes place by combining the data of the digital road map, an in 7 Priority context VK, the merged features of road users V1 to Vn and the risk assessment RB of pairs of road users V1 to Vn.

In the risk assessment RB, there is a risk of collision for each situation hypothesis DA and / or consequential collision risk determined using logical context rules. The context rules include, inter alia, priority and traffic rules, speed limits, display conditions of traffic lights and variable traffic signs and the topography and topology of the corresponding digital road map. The hypotheses of the situation are hierarchically determined on the basis of the driver's intentions FA as a function of the respective risk of collision DA and / or consequential risk of collision in an in 7 and 9 ordered priority list PL ordered.

The data of the digital road map include a topography of the shear and / or Ausscherspurs and a topology of these in relation to the lanes FS1 to FS4 of the road.

The priority context VK results from data indicating information about a type of lane markers, i. H. a continuous or interrupted training as well as a course of this and a derivable therefrom training the tracks as Einscher- and / or Ausscherspuren include. Furthermore, the priority context VK results from registered traffic light systems, a right of way regulation and traffic signs located in the surroundings.

As a result of the risk assessment RB, a collision risk DA is determined from a development of a conflict between at least two road users V1 to Vn. Conflicts are either potential conflicts or real conflicts. Another result of risk assessment RB are input signals ES for an in 9 illustrated roadway planning Y with the smallest risk of collision DA.

Due to a probabilistic combination of at least part of the aforementioned features, the robustness of the probabilistic hypothesis estimation is ensured. These features include in particular the relative, in particular speed-dependent, distance and a relative orientation of the own vehicle to at least one other road user V1 to Vn, the object-centered dynamic occupancy grid per road user V1 to Vn in the vicinity of the own vehicle and the absolute lateral distance and / or the orientation of all road users V1 to Vn including the own vehicle to the existing, for example by means of a camera detected lane markers or virtual lane markers. Furthermore, the features include the context information, including the topography and topology data and the traffic rules, traffic signs and the movement states BZ1 to BZn, the control states SZ1 to SZn, the driver activity states FAZ1 to FAZn and the event messages EN.

7 shows a method sequence for determining a collision probability between two road users V1 to Vn and an output of information Inf in a first escalation level and a first warning W1 in a second escalation level.

From the positions POS1 to POSn of the road users V1 to Vn, the movement states BZ1 to BZn and further input signals ES, the driver's intentions FA, also referred to as maneuver intentions, are first of all determined. The input signals ES include the in 5 shown variables for determining the driver's intent FA and cooperative perceptions, which result from a cross-perception, a vehicle-local perception as well as a self-localization.

Based on driver intentions FA1, FA2 of drivers of two road users V1 to Vn trained as vehicles and maneuver pairs MP with intersections and a context X (K1 to Ku) derived from the digital road map of the conflict areas K1 to Ku, a potential risk of collision DA and / or subsequent collision risk is interposed determined the two road users V1 to Vn.

This determination of the risk of collision DA and / or consequential risk of collision occurs for all competing situation hypotheses, wherein the situation hypotheses are hierarchically arranged on the basis of the driver's intentions FA and depending on the respective collision risk DA and / or consequential collision risk of the priority list PL.

If a potential risk of collision DA and / or consequential collision risk is determined, which exceeds a predetermined threshold value, the driver of the respective vehicle, ie one of the road users V1 to Vn, is reached before reaching a first distance to another road user V1 to Vn or at a predetermined first time duration until reaching the first distance in a first escalation stage at a first time optical, acoustic and / or haptic information Inf about the potential risk of collision DA and / or consequential risk of collision in the respective vehicle interior of trained as vehicles road users V1 to Vn issued to the driver. The first distance and the first time duration are preferably fixedly or as a function of a relative speed between the respective road users V1 to Vn and / or a coefficient of friction of the road user V1 to Vn to a pavement variably specified. In this case, the information Inf relating to the situation hypothesis or situation is output, which has the highest priority in the priority list PL. The first time is in particular more than 3 seconds before a possible collision K. The information Inf is thereby output in particular optically with a light green and transparent background and a voice output and serves to direct the attention of the driver in the direction of the location of the danger.

In particular, the information inf contains data about the size of the risk of collision DA and / or about the risk of subsequent collision.

Subsequently, a risk assessment RB is carried out on the basis of the potential risk of collision DA and / or subsequent risk of collision, of the priority context VK and on the basis of occupancy times t (K1 to Ku) of the conflict areas K1 to Ku. If a predetermined limit value is exceeded, in a second escalation stage at a second time following the first time, an optical, acoustic and / or haptic first warning W1 is output before a real risk of collision DA and / or subsequent collision risk for the road users V1 to Vn in the respective interior. The first warning W1 is sent to the driver of the respective vehicle, i. H. one of the road users V1 to Vn, issued before reaching a second distance to another road user V1 to Vn or at a predetermined second time period until reaching the second distance. The second distance and the second time duration are preferably fixedly or as a function of a relative speed between the relevant road users V1 to Vn and / or a coefficient of friction of the road user V1 to Vn set to a road surface variable. In this case, the first warning W1 is output with regard to the situation hypothesis or situation which has the highest priority in the priority list PL. The second time is in particular between 2 and 3 seconds before a possible collision K. The first warning W1 is output in particular optically with a yellow and transparent background and interrupted sound signals. Depending on the criticality, the optical output can be interruption-free or flashing.

The first warning W1 implies that a collision K is potentially possible and is determined solely from the knowledge of possible intersections between the trajectories or between the maneuver options of the road users V1 to Vn.

The illustrated structure enables a qualitative knowledge representation by the structure of the object-oriented Bayes network, by causality relations of the states, by an environmental context by means of the topology and topography of the digital map, by the priority context VK and by the dynamic behavior of the road users V1 to Vn a quantitative representation of the dependencies due to the determination of probabilities of collisions K between the road users V1 to Vn according to a strength of the relations between the road users V1 to Vn possible. At each time step, a new evaluation of the determined probabilities of potential and real risk of collision occurs. The update is performed by the inference algorithm, for example in the object-oriented Bayesian network, and is based on the locally measured and communicated data at each time step.

A driver state FZ and the risk assessment RB serve as filters for an output and intensity of the information Inf and warnings W1 to W3 in different escalation stages.

In 8th FIG. 3 illustrates a process flow of a method of assisting a driver in driving a four escalation vehicle.

9 shows the in 8th shown procedure and a linkage of this procedure with the in 7 illustrated procedure for implementing a method for assisting the driver when driving a vehicle. The vehicle is any road user V1 to V4 designed as a vehicle. Not only "4-way intersections" are considered, but also other topologies and topographies of the intersection with the corresponding vehicles V1 to Vn, which are assigned to the corresponding lanes (> 4).

The combination of the process sequences results in a common process sequence for determining the risk of collision DA and / or consequential risk of collision between two road users V1 to Vn, by means of which in different escalation stages information Inf, warnings W1 to W3 and an intervention in a longitudinal or lateral control of road users V1 to Vn depending on the risk of collision DA and / or consequential risk of collision is realized. The reference symbol p stands for the terms "positive" and "yes", the reference symbol n stands for "negative" and "no".

Up to the second escalation level, the information and warning concept follows 7 , then as shown below taking into account the driver state FZ, which is determined from an attention level, a degree of distraction, a degree of fatigue, stress and / or vital parameters of the respective driver by means of at least one optical detection unit, operator actions and / or other detection units.

Subsequently, the movement margins, the so-called "feel-safe zone", between the road users V1 to Vn determined, the determination, for example, according to the DE 10 2011 106 176 A1 and / or the not yet published DE 10 2012 005 272 he follows. Furthermore, possible control variables are determined and evaluated probabilistically in order to avoid a collision K between the road users V1 to Vn. These control variables include first control variables ICP1, which make it possible, without collision-free execution of an extreme maneuver EM, to drive past the respective other road user V1 to Vn, ie the danger vehicle from which the danger for the respective road user V1 to Vn originates. Under an extreme maneuver EM are understood, for example, full braking, maximum acceleration and a critical steering maneuver.

If such first control variables ICP1 are present, a second visual, acoustic and / or haptic warning W2 is output to the driver in a third escalation stage, which indicates to the driver that a real risk of collision DA and / or subsequent collision danger with a range of motion and available control variables ICP1 collision-free onward journey. The second warning W2 is given to the driver of the respective vehicle, i. H. one of the road users V1 to Vn, issued before reaching a third distance to another road user V1 to Vn or at a predetermined third time period until reaching the third distance. The third distance and the third time duration are preferably given fixedly or as a function of a relative speed between the relevant road users V1 to Vn and / or a coefficient of friction of the road user V1 to Vn to a road surface. In this case, the second warning W2 is output with regard to the situation hypothesis or situation which has the highest priority in the priority list PL. The second warning W2 takes place, in particular, 1 s to 2 s before the entry of the potential collision K. The second warning W2 is emitted in particular optically with a light-orange and transparent background and a haunting sound signal. Depending on the criticality, the optical output can be interruption-free or flashing.

If there is no correct action H of the driver on the second warning W2, but first control variables ICP1 are present for the collision-free bypassing of the other traffic participant V1 to Vn, an intervention in the driving dynamics of the vehicle's own traffic participant V1 to Vn takes place in such a way that the Road users V1 to Vn without extreme maneuver EM and without collision on the other road users V1 to Vn is passed. For this purpose, a longitudinal and / or lateral control of the respective vehicle is influenced in such a way that the vehicle is automatically slowed down, stopped, accelerated and / or steered and / or acceleration by the driver is prevented.

That is, the probabilistically evaluated control variables ICP1 represent the maneuver options still available to the driver. In the absence of a driver reaction, the maneuver avoiding the collision K is autonomous by means of a driver assistance device and, if first control variables ICP1 are available, for the occupants of the vehicle participants V1 to Vn comfortable and preferably carried out in the typical driving style of the driver.

If no first control variables ICP1 are present and if the driver does not act correctly to avoid the collision K, a third visual, acoustic and / or haptic warning W3 is issued before the driver in a fourth escalation stage before an acute danger of collision DA and / or after collision There is an automatic intervention in the driving dynamics of road users V1 to Vn under execution of an extreme maneuver EM. The third warning W3 is sent to the driver of the respective vehicle, i. H. one of the road users V1 to Vn, issued before reaching a fourth distance to another road users V1 to Vn or at a predetermined fourth time period until reaching the fourth distance. The fourth distance and the fourth time period are preferably fixedly or as a function of a relative speed between the respective road users V1 to Vn and / or a coefficient of friction of the road user V1 to Vn set to a pavement variable. In this case, the third warning W3 is output with regard to the situation hypothesis or situation which has the highest priority in the priority list PL. When determining the possible extreme maneuvers EM and their execution, a remaining time for braking, a remaining time for steering and a remaining time for acceleration or for kick-down are determined and taken into account. The third warning W3 takes place, in particular, less than 1 s before the entry of the potential collision K. The third warning W3 is emitted in particular optically with a bright red and transparent background and a haunting sound signal. Depending on the criticality, the optical output can be interruption-free or flashing.

As a result, the collision K is prevented or at least its consequences are minimized and the roadway planning Y with the smallest risk of collision DA and / or subsequent collision risk can be performed.

Due to the consideration of the degree of attention and the vital data of the driver in the assessment of the range of motion unnecessary and irritating driver warnings are avoided.

All information Inf and warnings W1 to W3 are detected in the vehicle interior from a direction of a relative position of the collision-dangerous road user V1 to Vn, i. H. of the road user V1 to Vn from which the danger emanates issued. For example, depending on the position of the collision-dangerous traffic participant V1 to Vn, an acoustic driver warning is emitted in front of the relevant road user V1 to Vn on the basis of signals from a left or right A-pillar. The acoustic signals are amplified by optical signals.

The optical signals are generated, for example, in case of front or rear approaching longitudinal traffic in a blind spot display or by means of a transparent coloring of edges of an inner rearview mirror. The danger from a rear or front left or right direction occurs, for example, even if a dangerous or vulnerable road users V1 to Vn starting from a separate vehicle, which is also part of road users V1 to Vn, on the left or right side of Roadway is located. In the case of traffic arising from oncoming traffic, the optical signals are emitted, in particular, at an upper edge of a windshield.

The possibility of such a detection of the dangers and the consequent position-related output of the information Inf and warnings W1 to W3 results in a particularly advantageous manner from the fact that the environmental information of the road users V1 to Vn are detected in a detection range of 360 °.

The limits of the human and physically feasible actions H of the driver before an autonomous intervention in the driving dynamics are set by the available range of motion and if these limits are exceeded, by a criticality determination of a situation between the road users V1 to Vn.

The determination of the criticality for a situation takes place as a function of determined time measurements, the time measurements being a time until the collision K, a time until a possible braking, a time until a possible steering intervention and / or a time until a possible kick. Down include an accelerator pedal.

The criticality correlates with the priority of the respective situation hypothesis in the priority list PL. This means in particular that with high criticality the respective situation hypothesis has a high priority in the priority list PL.

The hierarchical priority list PL of the hazards sets the priorities in the situation and assists the driver by issuing the information Inf and the warnings W1 to W3 and serves to guide attention while providing a basis for autonomous engagement in the longitudinal and / or or lateral control of the vehicle and for determining the intensity of information Inf and warning W1 to W3. The intensity is set by variation, in particular of audio signals combined with optical signals. The optical signals are generated, for example, by means of transparent organic light emitting diodes, conventional light emitting diodes and / or fiber optics. The optical signals are output with or without directions, which indicate the location of the risk of collision DA.

In particular, the driver of the respective vehicle probabilistically recognized driver intentions of the other drivers by means of an arrow display can be issued. This is particularly advantageous in the case of covered road users V1 to Vn, in particular in the case of a driver state FZ of the system vehicle at a low level.

Thus, the road users V1 to Vn can easily understand the dangers and gain confidence in the process of assistance.

In particular, the determination and a quantitative assessment of the criticality are carried out in the probabilistic probabilistic method as well as in methods with knowledge representation and in so-called black box methods.

The determination and the quantitative assessment of the criticality as well as the determination of the priorities can take place with different procedures, whereby the derivation of priority lists PL requires procedures, which provide both a qualitative and quantitative knowledge representation for the evaluation of a situation with all involved road users V1 to Vn enable. The result triggers the corresponding interventions in the longitudinal and / or lateral control as well as the output of the information Inf and warnings W1 to W3 in the four escalation stages. This capability of the situation analysis is achieved in particular by probability-based methods with a structured representation of the already described dependencies between the road users V1 to Vn and between the features that these road users V1 to V1n in a traffic situation.

Probability-based methods include the object-oriented Bayesian networks (OOBN for short), which use directional acyclic graphs with causal dependencies between the sizes, weighted by probabilities. The object-oriented Bayesian networks allow a combination of insecure and heterogeneous information / data sources and can also deal with missing evidences.

The probability-based methods also include object-oriented influence diagrams (OOID), which represent an extension of the object-oriented Bayesian networks with cost functions and time functions, as well as so-called Hidden Markov Models (HMM), which is an extension of the object-oriented Bayes Networks as dynamic prediction models, assuming that the future is independent of the past, when the sizes are known in the present.

The so-called Dempster Shafer Theory (DST) - also known as evidence theory - is also used as a probability-based method. The Dempster-Shafer theory is used to assemble information from different sources into one overall statement, taking into account credibility of these sources in the calculation, similar to the object-oriented Bayesian networks.

The probability-based methods also use so-called fuzzy logic, whereby fuzzy logic membership degrees are interpreted as probabilities. The fuzzy logic represents the dependencies. Also so-called adaptive neuro-fuzzy systems are used, which combine the fuzzy logic with the learning ability of neural networks.

Furthermore, decision trees are used, which are designed as ordered or directed trees and serve to represent decision rules. The decision trees illustrate hierarchically consecutive events and / or decisions that are strongly context dependent.

A combination of such decision trees with neural networks leads to tree-based neural networks (abbreviated to TBNN), which are also used as probability-based methods. In this combination, inefficient branches of the decision tree are replaced by neural networks to achieve a higher classification quality. Thus, the advantages of both classification methods are used, which consist in that only a small number of training data for the trees required for induction and at the same time an accurate classification is possible.

In addition, support vector machines (SVM) and the kernel method are used as classifiers, which contain a set of objects, ie. H. of road users V1 to Vn, subdivided into classes so that the widest possible area remains free of objects around the class boundaries.

Kalman filters and extensions are also used as a linear variant of Bayesian networks.

By contrast, qualitative methods are only suitable for the presentation of knowledge and are not suitable as the first choice for creating the priority list PL due to the lack of efficient search algorithms. The qualitative methods include an ontology that represents a formal network of information with logical relations. The possibility of establishing relations via relations and rules is relatively seldom used in practice because of its complexity, although ontology in particular differs from other conceptual systems. Also known as XML-structured vehicle driver environment event trees are among the qualitative procedures. These are more suitable as well-structured knowledge databases for the above-mentioned probability-based methods with knowledge representation, where only the dependencies between the input variables are represented, but do not include the strength of these dependencies. This quantification is necessary for the inference algorithms to calculate a hierarchically ordered priority list PL and to update each time step.

For such quantifications of, in particular, statically remaining features of the objects, in particular pure quantifications, the black box methods are suitable, whereby scaling effects can still be made recognizable by means of these methods. The black box techniques include so-called neural networks (NN for short), Bayesian neural networks, which are Bayesian probability extended neural networks, and include approximation algorithms such as polynomial, least-squares, spline, and others.

However, these black box methods are not suitable for dynamically changing situations that do not always routinely follow the same patterns. They require elaborate and renewed learning procedures when certain features of the situation have changed quantitatively. Thus, they are not suitable as so-called online methods. For training and learning is a big one Amount of data required, which are characteristic for the object under classification. There is also the danger of so-called overtraining, ie they can not generalize when a new similar situation arrives.

Suitable applications for the black box method are the detection of traffic signs, the detection of phases of a traffic light, the detection deviations from the normal operation of a system or component, the classification of typical turn trajectories in existing reference database and associated maneuver intention detection and the Detection of deviations from the driver's normal behavior. Thus, the black box methods are suitable for the processing of input data and for the information level.

a

Beschleunigung

B1 bis B4

Gefahrenbereich für die Bewegungsspielraum

BZ1 bis BZn

Bewegungszustand

DA

Kollisionsgefahr

dG

dynamisches Gitter

EA

Einscher- und/oder Ausschervorgang bzw. Einfädeln- und/oder Ausfädelvorgang

EN

Ereignisnachricht

EM

Extremmanöver

ES

Eingangssignal

FA, FA1, FA2

Fahrerabsicht

FAZ1 bis FAZn

Fahreraktivitätszustand

FS1 bis FS4

Fahrspur

FZ

Fahrerzustand

G

Geradausfahrvorgang

H

Handlung

I1

Information

Inf

Information

ICP1

Steuergröße oder Steuerprofil des Fahrers

K1 bis Ku

Konfliktbereich

K

Kollision

I

Lenkrate

M1, M2

Menge

MP

Manöverpaar

n

negativ/nein

O1_rel

erste relative Orientierung

O2_rel

zweite relative Orientierung

p

positiv/ja

P(C)

Gefahrenwahrscheinlichkeit für die Bewegungsspielraum

PL

Prioritätsliste

POS1 bis POSn

Position

RB

Risikobewertung

S

Spurwechselvorgang

SS1 bis SSu

Straßensegment

SZ1 bis SZn

Steuerzustand

T1₁ bis Tm₁

Bewegungshypothese-Trajektorie, m₁-Anzahl der mögliche Bewegungshypothesen für Verkehrsteilnehmer V1

T1₂ bis Tn₂

Bewegungshypothese-Trajektorie, n₂-Anzahl der mögliche Bewegungshypothesen für Verkehrsteilnehmer V2

t(K1 bis Ku)

Belegungszeit des Konfliktbereichs K1 bis Ku

MT1 bis MT12

Manöver-Track

U

Überholvorgang

V1 bis Vn

Verkehrsteilnehmer

v1_rel

erste Relativgeschwindigkeit

v2_rel

zweite Relativgeschwindigkeit

VK

Vorrangkontext

W1 bis W3

Warnung

X(K1 bis Ku)

Kontext Konfliktbereich

Y

Fahrbahnplanung

In contrast, the above-mentioned probability-based methods with the knowledge representation already contain four escalation levels for the structured construction of the model and thus offer a compact and efficient modeling approach.

a

acceleration

B1 to B4

Danger area for the range of motion

BZ1 to BZn

moving state

THERE

risk of collision

dG

dynamic grid

EA

Einscher- and / or Ausschervorgang or threading and / or Ausfädelvorgang

EN

event message

EM

extreme maneuvers

IT

input

FA, FA1, FA2

driver intent

FAZ1 to FAZn

Driver activity state

FS1 to FS4

lane

FZ

driver condition

G

Geradausfahrvorgang

H

action

I1

information

Inf

information

ICP1

Control variable or control profile of the driver

K1 to Ku

conflict area

K

collision

I

steering rate

M1, M2

amount

MP

maneuvers couple

n

negative / no

O1 _rel

first relative orientation

O2 _rel

second relative orientation

p

positive / yes

P (C)

Hazard probability for the range of motion

PL

priority list

POS1 to POSn

position

RB

risk assessment

S

Lane-changing operation

SS1 to SSu

road segment

SZ1 to SZn

control state

T1 ₁ to Tm ₁

Motion hypothesis trajectory, m ₁ number of possible motion hypotheses for road users V1

T1 ₂ to Tn ₂

Motion hypothesis trajectory, n ₂ number of possible motion hypotheses for road users V2

t (K1 to Ku)

Occupancy time of conflict area K1 to Ku

MT1 to MT12

Maneuver Track

U

overtaking

V1 to Vn

road users

v1 _rel

first relative speed

v2 _rel

second relative speed

VK

priority context

W1 to W3

warning

X (K1 to Ku)

Context conflict area

Y

road planning

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

• DE 102011120117 [0002]
• DE 102011113019 [0033, 0043]
• DE 102012005272 [0040, 0043, 0052, 0125]
• DE 102011106176 A1 [0052, 0125]

Claims (10)

Method for assisting a driver in driving a vehicle, wherein driver information is issued depending on a predicted future potential, real or acute risk of collision (DA) and / or subsequent collision risk between the vehicle and other road users (V1 to Vn) within the vehicle, characterized that - maneuver intentions (TM1 to TMn) are probabilistically identified on the basis of characteristics which are characteristic of a position and a movement state of the vehicle and the other road users (V1 to Vn), - a risk of collision based on intersections of detected pairs of maneuver intentions (TM1 to TMn) of the traffic participants (V1 to Vn) located in relative movement with respect to one another and a corresponding spatial occupancy of conflict areas, - a probabilistic risk assessment (RB) of a respective real or acute risk of collision (DA) and / or consequential collision sion risk based on the allocation of conflict areas in space-time and in combination with a logical context rule, including a priority rule. A plurality of mobile movement hypothesis trajectories of the vehicle and of the other road users (V1 to Vn) are predicted from these on the basis of control profiles, using the trajectories to determine a number of competing situation hypotheses between the vehicle and all other relevant road users (V1 to Vn), - An existing range of motion with appropriate probabilities for a real risk of collision with existing control variables for collision avoidance or an existing range of motion with appropriate probabilities for acute danger of collision and a need for extreme maneuvers for Kollisionsentschärfung be determined and the probabilistic results obtained in dependence of the respective risk of collision (DA ) and / or subsequent collision risk hierarchically in a priority list (PL), - the driver of the vehicle depending on the priority list (PL) and driving he presettings according to a time point of a desired system intervention and / or a determined driver status (FZ) of the driver in several escalation levels by means of the driver information is informed, warned and / or assisted by an automatic intervention in a longitudinal and / or lateral control of the vehicle, and An information (Inf), a warning (W1 to W3) and an automatic intervention are made for a situation which has a risk of collision (DA) and / or following collision danger with a highest priority in the priority list (PL) and a consequent highest criticality , A method according to claim 1, characterized in that for determining the situation hypotheses possible mutual intersections of the detected maneuver intentions (TM1 to TMn) of each other in relative motion road users (V1 to Vn) and the vehicle and based on the intersections in the conflict areas a potential risk of collision (K) is determined. A method according to claim 1 or 2, characterized in that for cognitive risk assessment in space-time, the real risk of collision (K) and / or consequential risk of collision based on the combination of the pairs of maneuver options between the vehicle and all relevant road users (V1 to Vn) together with the corresponding concurrent occupation of the conflict areas and the logical context of the digital map and the traffic rules is determined. A method according to claim 1, characterized in that possible collision-free movement hypothesis trajectory pairs of the vehicle and the road users (V1 to Vn) are predicted and evaluated for real risk of collision and in the evaluation of respective movement margins between the vehicle and the road users (V1 to Vn) be determined and depending on a size of the respective range of motion, the hazard probability (P (C)) is determined. Method according to one of the preceding claims, characterized in that the criticality is determined on the basis of determined time measurements, wherein the time measurements include a period of time until the collision (K), a time until a possible braking, a time until a possible steering intervention and / or a period of time to a possible kick-down of an accelerator pedal include. Method according to one of the preceding claims, characterized in that the information (Inf), the warning (W1 to W3) and the automatic intervention are performed in four escalation stages, wherein - in a first escalation stage at a first time an optical, acoustic and / or or haptic information (Inf) about a potential risk of collision (DA) and / or subsequent collision risk of the vehicle with at least one of the road users (V1 to Vn) is output, - in a second escalation at a second time following the first time an optical, audible and / or haptic first warning (W1) before a real collision hazard (DA ) and / or consequential risk of collision of the vehicle with at least one of the road users (V1 to Vn) is output, - in a third escalation stage at a third time following the second time an optical, acoustic and / or haptic second warning (W2) before a real Risk of collision (DA) and / or consequential risk of collision of the vehicle with at least one of the road users (V1 to Vn) and without need for extreme maneuvers is output and - in a fourth escalation stage at a fourth time following the third time a visual, audible and / or haptic third warning (W3) with an automatic system intervention in a longitudinal and / or transverse control of the vehicle to protect against an acute danger of collision (DA) and / or consequential risk of collision of the vehicle with at least one of the road users (V1 to Vn) is output. Method according to one of the preceding claims, characterized in that for determining the situation hypotheses by means of an object-oriented Bayes network lateral evidence, trajectories of the vehicle and road users (V1 to Vn), object-oriented dynamic grid (dG) in the conflict areas around the road users around and contextual information probabilistically combined to the intended maneuver recognition. Method according to claim 7, characterized in that as context information From a digital road map of a road segment (SS1 to SSu) on which the vehicle is located, - From signals from vehicle-mounted sensors of the vehicle and From data of a vehicle-to-vehicle communication and / or a vehicle-to-infrastructure communication between the vehicle and the other road users (V1 to Vn) and / or an infrastructure A movement state (BZ1 to BZn) of the vehicle and the road users (V1 to Vn), a control state (SZ1 to SZn) of the vehicle and the road users (V1 to Vn), a driver activity state (FAZ1 to FAZn) and event messages (EN) become. A method according to claim 7 or 8, characterized in that when determining the situation hypotheses, a distance, a relative orientation and a relative positioning of the vehicle and the road users (V1 to Vn) are determined. Method according to one of the preceding claims, characterized in that the information (Inf) and / or the warning (W1 to W3) in the event of collision risk (DA) and / or consequential risk of collision depending on a relative position of the at least one road user (V1 to Vn) is outputted to the vehicle in the interior of the vehicle at an indoor position, which is located by the driver substantially in a direction in which the at least one road user (V1 to Vn) is relative to the first vehicle.

Images